<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Elvan Aydemir on Medium]]></title>
        <description><![CDATA[Stories by Elvan Aydemir on Medium]]></description>
        <link>https://medium.com/@l1-aeg?source=rss-c8c90cb6e013------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*ieeSv3QsC-GdXVjs_zkQNQ.png</url>
            <title>Stories by Elvan Aydemir on Medium</title>
            <link>https://medium.com/@l1-aeg?source=rss-c8c90cb6e013------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 03:08:35 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@l1-aeg/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Burden of Time]]></title>
            <link>https://l1-aeg.medium.com/burden-of-time-d23510909d21?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/d23510909d21</guid>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Sun, 04 Aug 2024 20:12:57 GMT</pubDate>
            <atom:updated>2024-08-04T20:12:57.319Z</atom:updated>
            <content:encoded><![CDATA[<p>Due to a series of decisions I made, and some unexpected bumps along the way, I recently found myself with a lot more time in my hands than usual. Naturally, as one’s life gets less busy, one’s mind tends to get more so. Needless to say, I need an outlet, hence the blog.</p><p>Most recent thoughts of mine are all about time, the past and the future. My “now” is uneventful, which is a blessing on its own, but that means there isn’t much to think about that. I suppose it is a great opportunity to practice mindfulness and staying in the now, but frankly I have been failing.</p><p>I guess it is expected that time overwhelms you when you have too much of it, much like anything else. I wonder how people go through their lives, not being petrified about all the inifinite probabilities of how their lives can go, all the decisions that they would have to make which will branch out to even more and more decisions. Is this why people live almost scripted lives? While truly unique in all their details, essentially the same blueprint? Grow up, get a job, get a family, go through the motions of what is supposed to be a socially prestigious life? I used to think I was so above all of it. Or not really above, just <em>outside</em>.</p><p>Now I am wondering, whether following a blueprint was the right approach after all? Making the conventional choice? What is so wrong with conventional? It <em>is</em> convention for a reason. The path of least resistence. Why do I keep picking the path of non-least resistence? Why do I keep diverging from the well-treaded path? Often times I get stuck in bushes and it is unpleasant AF and then it gets dark and then I am hungry and cranky. Left wanting.</p><p>That’s the problem I guess. <em>Wanting.</em> Maybe the collective wisdom of conventional people knows better. Maybe I shouldn’t want something different. Then I wouldn’t have to make any decisions. Maybe then <em>my brain will stop hurting.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d23510909d21" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A 2.5 Year Reflection of My Own Climbing]]></title>
            <link>https://l1-aeg.medium.com/a-2-5-year-reflection-of-my-own-climbing-200443cb9e33?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/200443cb9e33</guid>
            <category><![CDATA[sports]]></category>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[sports-psychology]]></category>
            <category><![CDATA[climbing]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Sun, 01 Oct 2023 09:12:53 GMT</pubDate>
            <atom:updated>2023-10-01T09:14:03.908Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/900/0*vjA8V5tuHveR4sXs.png" /><figcaption>Cryptic knot image to represent how tangled the inside of my brain is</figcaption></figure><p>Disclaimer: Following post is more of a ramble than a coherent collection of thoughts because I don’t really know where to begin or end writing.</p><p>I am an overthinker and I have been one my whole life. I stress over every possible combination of actions and consequences I can think of when it comes to decision making. Some of my personal decisions may have been perceived as impulsive (by my immediate family mostly), and they largely were, but not because I wasn’t thinking about the way they would play out but rather I ended up getting overwhelmed by the barrage of possibilities and said “Fuck it.” and picked something in the end.</p><p>You may be wondering how this is related to my climbing. Well, as an overthinker, I am fairly predisposed to being a certain type of climber: A project climber. And over the past 2.5 years that I have been climbing, this is what I have been, a project climber.</p><p>As far as climbing goes, 2.5 years is a blink of an eye. Most climbers climb for decades, although I am convinced this stat is skewed due to survivor’s bias as many quit within the first year. Regardless, the ones who stay in the sport stay for a long time.</p><p>In those 2.5 years, I managed to climb some hard routes for the climbing tenure I had, largely because of my overthinking and (I have to admit) because of my rather large insecurities (which affect everything in my life but that is a topic for another time). I felt (and still do feel) comfortable sitting on a rope, obssessing over a 2 cm difference between two possible footholds, calculating whether the higher one which is way worse than the lower one would give me a marginal improvement in making the next big move to the hold. I could (and still can) stay on a rope for 2 hours, trying the same move with those two different footholds, stepping on them in different angles, shifting my hips to feel if there is a better balance point, dropping the other leg, dropping the other hand etc. Just you name it, every combination you can think of, I am inclined to try until I am <em>certain</em> I cannot improve the move given my current abilities. This obssessing over is als futher fueled by the said insecurities because (quoting the voice inside my head) “If I can’t do this climb/move, what good am I?”. Yeah, pretty fucked up. But this resulted in a few things:</p><p>1- My physical climbing abilities have gotten better very fast because trying hard moves and really <em>intentionally</em> learning from them just expands your climbing abilities. You get stronger and you are forced to understand why a move works and why it doesn’t, and you are forced to learn new techniques quite fast if you want to do that move</p><p>2- I can try and maybe climb routes above my pay grade, and when I send them, I look kinda like I know what I am doing (I really don’t know what I am doing)</p><p>3- I have gotten so comfortable working out every single minute detail that I have no idea how to “just climb” things</p><p>4- I have developed an even more lopsided set of mental abilities. I am good at being patient with projects, I can persevere and keep going back to them even when it seems hopeless but when it comes to flashing/onsighting I have no mental tools and techniques to just try a sequence see if it works or not. As soon as something gets even a little complicated while onsighting, I go to project mode and instantly want to explore all the possible ways to climb that section instead of just choosing one and committing.</p><p>5- As much as I hate to admit this, I am afraid of failing to flash/onsight “easier” routes, because it is closely related to my self-worth and if I can’t onsight a 6b while climbing 7b+, my brain tells me I am not a real climber and I get MASSIVE impostor syndrome. It sucks big time and makes me feel like utter garbage, so I avoid even trying. Bleh.</p><p>Above all of the issues and the triumphs, I simply love climbing. There has not been a single day out climbing (no matter how difficult it was mentally/physically) where at the end of the day I didn’t feel like it was an amazing day. Not going to lie, whether I send or not matters to me immensely, but I am finding more and more that it is less and less tied to my self-worth. At some point in time, not sure when, I started to embrace the idea of failing, regardless of the type of climbing (onsight/project) or the grade of the climb.</p><p>Funnily enough, now that I am slightly more okay with the idea of failing/not sending (not even remotely close to being completely ok with it but getting there) and having it be less damaging to my self-worth, I am enjoying climbing a lot more. My goal is no longer to “tick the climb” or “climb harder”. I don’t choose routes based on their grade as much, I choose routes based on first and foremost whether I will enjoy the climbing. Don’t get me wrong, I still enjoy doing moves that are hard for me, that’s just what I like (hello moonboard, I love you). That generally means I gravitate towards climbs that have hard (for me) moves. But if two routes are close in terms of their challenge for me, I will choose the nicer one even if it is less “prestigious” in terms of grade.</p><p>I have also been finding that lots of “easier” climbs have moves that are hard for me for one reason or the other. And I noticed recently that this annoys me if the climb is “supposed to be” easy. That is such a dumb thought. If the move is hard, it is hard, and if that is what I enjoy, I should be able to enjoy it regardless of the grade of the route. But that’s ego for you. My ego is bruised when I struggle on a route that is supposed to be “easy” for me. And it is even more bruised when someone says “This should be easy for you” because then I am not only failing to fulfill my own expectations, I am also failing to live up to my image, whatever that means. This is such a dumb train of thought. If I am finding a move hard, there is a reason for this, and there is something to be learned. I can’t control other people’s expectations, that’s their problem really, and I don’t even know why I think other people expect anything tbh, they probably don’t, but my own expectations are somewhat in my own control. Once I noticed all of these complex, ego-driven subconscious thought processes, and realised that actually, what I care about the most is enjoying my climbing and climbing <em>well, </em>I started enjoying struggling on easier routes quite a bit. It is an opportunity to learn quickly whereas otherwise I would be spending hours just to even <em>begin</em> trying to learn on a route that is truly at my limit.</p><p>So, enough rambling. Bottomline is, my goals have shifted. For whatever reason for the past 2 years or so, I had a chip on my shoulder, trying to prove something to myself and to other people I suppose by climbing “hard” grades that are supposed to be out of my reach. This chip has started to fall a bit. Now I don’t really have any quantifiable goals (such as climb X amount of grade Y and climb Z grade by date D). My goal is to be proud of my climbing and get more confident in my climbing abilities. Climbing is a sport where you never really stop learning, there is no end. I will always have weaknesses, have things to improve on. But this is what makes it endless and life-long. And that’s what I love.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=200443cb9e33" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Roe v. Wade Overturned And We Should All Be Furious]]></title>
            <link>https://l1-aeg.medium.com/roe-v-wade-overturned-and-we-should-all-be-furious-1993055be315?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/1993055be315</guid>
            <category><![CDATA[politics]]></category>
            <category><![CDATA[abortion]]></category>
            <category><![CDATA[women]]></category>
            <category><![CDATA[womens-health]]></category>
            <category><![CDATA[womens-rights]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Sat, 25 Jun 2022 16:54:11 GMT</pubDate>
            <atom:updated>2022-06-25T16:54:11.368Z</atom:updated>
            <content:encoded><![CDATA[<p>It doesn’t matter whether you are from US or no.</p><p>I can’t be bothered with looking for a catchy title for this. This will be a raw post.</p><p>I don’t normally post about world politics although I am fairly political (I simply prefer to discuss in person). But the decision to overturn Roe v. Wade, effectively allowing individual states to block women from accessing a vital operation to terminate unviable/unwanted pregnancies is infuriating. This decision will cost the lives of women and here are the names of the people responsible for it. Remember them:</p><ul><li>Clearance Thomas</li><li>Samuel A. Alito, Jr.</li><li>Neil M. Gorsuch</li><li>Brett M. Kavanaugh</li><li>Amy Coney Barrett</li></ul><p>In addition to this information, let me just say the rhetoric adopted by people protesting against this that goes like “what about teenage girls who are victims of rape??”, while a valid question, is not doing anything but reinforcing the belief that abortion is inherently wrong BUT may be allowed in certain extreme cases. No. Simply no. The bare minimum valid reason for terminating a pregnancy is that the pregnant party wants to terminate the pregnancy. That’s it. That’s all the reason you need. I won’t go into the biological states of pregnancy or whatever, that has already been established.</p><p>1- I get that some doctors may not want to perform this operation. While questionable, it is also understandable. Goes to personal values.</p><p>2- I also get that some women, in case they get an unwanted pregnancy, may decide to see to term (despite doubts and whatever) because they don’t believe that termination is the right thing for them. Again, goes to personal values.</p><p>However, UNDER NO CIRCUMSTANCE the people described above should be able to decide on behalf of the MASSES of women who do not share these ideas and would like to terminate. The AUDACITY to try to dictate how other people should govern their bodies is criminal, and we should all be furious. And making both seeking and performing abortions criminal is outright murder. There is already blood on the hands of these people. Remember them.</p><p>And let’s hope that this doesn’t set a precedent for the rest of the world.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1993055be315" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Dr. Stange and The Tale of Two Adults Missing What They Never Had]]></title>
            <link>https://l1-aeg.medium.com/dr-stange-and-the-tale-of-two-adults-missing-what-they-never-had-fee0db392524?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/fee0db392524</guid>
            <category><![CDATA[movie-review]]></category>
            <category><![CDATA[marvel]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Mon, 09 May 2022 11:59:43 GMT</pubDate>
            <atom:updated>2022-05-09T12:07:43.138Z</atom:updated>
            <content:encoded><![CDATA[<p>“Are you happy Steven?”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/540/0*RrctVlxbdXqq6tj3" /></figure><p>This is a review of the recent Marvel movie, Dr. Strange. It is not a regular review, it is one where project my issues onto the movie way too much but I think it is still pretty accurate. TLDR 8/10.</p><p>Disclaimer: Contains spoilers.</p><p>Seemingly, this movie is the classic “Two tortured souls, one becomes a hero and saves the world and the kid, the other one almost destroys it but good prevails in the end so does the right thing.” plot. Except it is not. It is not a superhero movie at all. There are all sorts of superpowers flying around but it is about two adults who have achieved godhood and yet failed to be happy.</p><p>“Are you happy Steven?” this sentence is the centerpiece of the movie. At the beginning of the movie, we see Dr. Steven Strange getting ready by using his magic to put on a suit for the wedding of the woman he loved/loves, with another man. He puts on his very broken wristwatch and goes to the wedding. Eventually, the bride asks him this question which he promptly lies to.</p><p>Wanda, the Scarlet Witch, repeatedly dreams of a family she never had. Tucking in her children that never existed, over and over again in the same dream every night.</p><p>Both these characters are portrayed as nothing more than humans, despite their obvious reality-altering powers. Dr. Strange is the most powerful sorcerer there is, in all multiverses, and yet in all of them he is searching for a way to reconcile with something he has lost a long time ago. Despite all his achievements and status, he is not happy yet he manages to go about his stuff with the occasional destructive stint. He will use his magic for literally any occasion, even in forbidden ways, but he won’t fix his goddamn broken wristwatch with it. His only link to what he lost.</p><p>Scarlet Witch is a reality-altering being that has incredible power and yet all she wants, despite the shrine made in her name and the power to destroy realms, is to be able to tuck her kids in. The kids she never had. The kids she had in every other reality but hers. “I am meant to rule over the universe.” she says “This is a throne, and I don’t want it.”.</p><p>This movie came on top of the numerous conversations I have been having with people about identity and accomplishments. So obviously the lens I had on when I walked into the theater was that lens. It is incredibly common among people, all adults really, to associate their self-worth and identity with what they have been able to accomplish. We strive and struggle for some imaginary achievement that will eventually make us say “That’s it. I am complete now.” Somehow this never seems to happen and we continue to ache and long for things we never had. Things we probably don’t even know we are missing.</p><p>We see both characters in the movie longing for what never was, which was never an accomplishment but love. Which is a cheesy plot in all honesty, but then again, it is a mainstream movie. What makes this good is that the “happy ending” is not about reconciliation this time, it is about moving on and letting go. The good doctor doesn’t get the girl and Wanda doesn’t get her children. Dr. Strange finally accepts Wong as the sorcerer supreme despite Wong not being his equal in the craft. But he is the superior man in terms of enlightenment. Wanda walks away from her children in another realm and leaves them with their actual mother who is an alternate self of hers and destroys her throne. We finally see Dr. Strange mending his broken watch, not with his magic but with traditional methods, almost with compassion, cherishing the memories but moving on nonetheless.</p><p>There is really no “A-ha” moment to this particular read of the movie. It is just a reflection of the Marvel cinematic universe audience consisting of as many (or maybe even more) adults as kids. The younger audience will love how action-packed the movie is, older audience too I think. Some of the members of the audience will take other things from the madness in it. Graphics are absolutely stunning as well. All in all, I enjoyed this immensely, and would recommend.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fee0db392524" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Climate Anxiety, Winter Sports and Self-Image]]></title>
            <link>https://l1-aeg.medium.com/climate-anxiety-winter-sports-and-self-image-c779a2e5e2fc?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/c779a2e5e2fc</guid>
            <category><![CDATA[climate-change]]></category>
            <category><![CDATA[snowboarding]]></category>
            <category><![CDATA[skiing]]></category>
            <category><![CDATA[self-image]]></category>
            <category><![CDATA[winter]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Tue, 15 Feb 2022 07:25:31 GMT</pubDate>
            <atom:updated>2022-02-15T07:25:31.311Z</atom:updated>
            <content:encoded><![CDATA[<p>Looks like I have thrown a whole bunch of big words in the title but I believe I have a very common story a lot of people can relate to, about a season-ending injury and insecurities that follow afterwards.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-veQRdPEoOPGmBkZ.jpg" /><figcaption>Source: <a href="https://www.mentalfloss.com/article/521061/climate-change-moving-so-fast-northern-alaska-its-confusing-computers">https://www.mentalfloss.com/article/521061/climate-change-moving-so-fast-northern-alaska-its-confusing-computers</a></figcaption></figure><p>For the longest time, I complained to my closest friends about not having an identity besides being a data scientist. “I am not really a person, I am a data scientist and if I don’t have that I don’t know who I am.” was a common sentence I would just put out there. Of course, they replied with all the other things that I had going on, none of them really registered.</p><p>That is until I (re)discovered the stuff that I used to do, and some new things. One of those things was skiing. I have been a skier since I was 5. I basically grew up on skis. Once upon a time, I even used to be good at it. For the past few seasons, I have been gradually building those skills back up. This season, I pushed hard on learning technique, taking several private lessons to comfortably ski down any groomed slope and some freeride terrain (in which I am still absolute garbage by any reasonable standard, although I can make my way down somehow).</p><p>One of the reasons why I was pushing so hard this season is because I don’t know how much longer we will continue to have winters. Every time I go on a lift, I cannot help but notice the rocks that used to be covered with snow are more and more naked. The towns just below the ski zones have little to no snow. And when we get snow, it is either a complete pour over a day or two, or we don’t get snow at all. Here in Sofia, Bulgaria we even had a day in February with 15 degrees Celsius temperatures. Fifteen. Degrees. Celcius. Going out in a leather jacket and <em>still being warm</em> was a terrifying feeling in February.</p><p>Then the next day I was on the slopes so early that lifts were not working and I walked up and down to ride, simply because I was terrified it would be gone before I knew it. S-tier anxiety right there.</p><p>The early morning was fine, and then it wasn’t. I took a bad fall. Torn meniscus. Season likely over unless my doctor clears me for riding in braces. I had started compiling my touring gear, so I could continue to ride in high mountains post-season (because, again, I am not sure how much longer we’ll have this). All gone in an instant. And I immediately regressed to an identity crisis.</p><p>I am not quite sure why I keep attaching my self-image to accomplishments or ‘being good at something’, whatever that may be. This is potentially a topic for long and frequent therapy sessions but the issue, in a nutshell, is this:</p><ol><li>We are facing a serious threat from climate change to life as we know it. People think COVID-19 devastated our lifestyle. I agree. But I also maintain that it was so apparent only because it was so sudden. Climate change will have a far deeper impact on everything we can and cannot do. COVID-19 will eventually be taken under control, we are getting there. If climate change gets out of hand, and it is very close to being so, there is absolutely nothing we can do to ‘fix’ it. Climate change has been gradually wrecking everything for decades now, and people don’t pay attention to it simply because it is gradual, unlike COVID.</li><li>I don’t really care about much, but I care a lot and very deeply about the stuff that I care about. So I have very severe reactions to being torn from those things. Hence the identity crisis. But, if climate change persists as is, I won’t be the only one having this issue. We will look back to these days in the future and realise we are so far from the person we were. This is obviously normal, people change and hopefully grow as time goes on. But some of the changes won’t be our choice. They would be forced on us by climate change, and I can almost guarantee those things won’t be in the way of growth. They would be in the way of compensating for the things we lost.</li></ol><p>Not quite sure why I am typing these down, I am a bit overwhelmed I guess. And very very upset, angry, scared. But there are things we can do in our private lives, they won’t make much difference unless large corporations and governments commit to climate action and <em>truly</em> deliver. The thing that would make a difference is voting with your wallet. Try not to buy from climate-damaging (or even climate-oblivious) companies. Banking is super important too, as banks fund most of the companies who do most climate damage. As of this post, I will do my best to change my banking.</p><p>I will just leave some resources below for people who are interested in exploring this, but please remember, imperfect climate action is better than none. You may not be able to do <em>all the things</em> to help with this, and as an individual your lifestyle impact will be small, but every bit counts. In any case, even if we fail in the end, we can say we did all we reasonably could. And that is some solace. About imperfect activism: <a href="http://raisingrevolutionaries.co.uk/2019/06/13/why-the-world-needs-more-imperfect-activists/">http://raisingrevolutionaries.co.uk/2019/06/13/why-the-world-needs-more-imperfect-activists/</a></p><ol><li>Protect our winters, targeted specifically to outdoors people <a href="https://protectourwinters.eu/">https://protectourwinters.eu/</a></li><li>General compilation of info on Wiki <a href="https://en.wikipedia.org/wiki/Individual_action_on_climate_change">https://en.wikipedia.org/wiki/Individual_action_on_climate_change</a></li><li>List of individual actions <a href="https://psychology.org.au/getmedia/2a2156ab-559c-4316-888a-a8cd82fcb780/101-things-you-can-do-climate-change_1.pdf">https://psychology.org.au/getmedia/2a2156ab-559c-4316-888a-a8cd82fcb780/101-things-you-can-do-climate-change_1.pdf</a></li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c779a2e5e2fc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Very Unpopular Review of Matrix: Resurrections]]></title>
            <link>https://l1-aeg.medium.com/a-very-unpopular-review-of-matrix-resurrections-e12962053d77?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/e12962053d77</guid>
            <category><![CDATA[movies]]></category>
            <category><![CDATA[matrix]]></category>
            <category><![CDATA[movie-review]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Thu, 06 Jan 2022 17:32:13 GMT</pubDate>
            <atom:updated>2022-01-06T18:35:12.320Z</atom:updated>
            <content:encoded><![CDATA[<p>It’s a good movie if you are not stuck in your teenage angst. I said what I said.</p><p>Contains spoilers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-la00q3E09l9awf-RrJaxg.jpeg" /><figcaption>Source: Matrix Resurrections</figcaption></figure><p>I went to see Matrix: Resurrections on the second day it aired in theaters and despite what most people who saw it on the first day said about the movie, I ended up really appreciating it. And I do love the original trilogy.</p><p>Subsequent conversations with people who are adamant about shitting on the movie all took the same turn so I decided to point out certain things people semi-deliberately insist on overlooking.</p><p>Initially, I jokingly said that the movie is perfect for people who love “self-deprecating humor, Glee, Sense8, contradictions, and machine-consciousness”. I think this sums up this blog post pretty well so unless you are interested in why I thought this was the case, you can stop reading. The review is going to be a lot more serious than this statement so prepared to be bored.</p><h3>Self-Deprecation and A &gt;20-Year-Later Look Back On Original Trilogy</h3><p>The most common critique (which has a lot of merits) of the new Matrix is that it is basically a compilation of the best scenes from the original trilogy because of a lack of ideas. While this has quite a lot of truth to it, I think those scenes are intended to highlight how over-the-top the original trilogy was. New “Morpheus” is more of a joker than the original one who was deeper and more solemn. So is agent Smith. To me, this felt like Lana Wachowski was not only mocking their own previous work but also the producing company that pushed a Matrix movie more than 20 years later. <a href="https://variety.com/2019/film/news/matrix-4-keanu-reeves-carrie-anne-moss-lana-wachowski-1203307955/">As Variety reports, WB have been trying to get back into Matrix universe for a while given the success of John Wick series.</a> This bit is a very speculative take that doesn’t have much basis for anything really, but I thoroughly enjoyed the fact that Wachowski had the courage to actively mock the Matrix and the cult of it. Now on to better reasons why the new movie is better than what people make it to be.</p><h3>A Critique on the Binary World(s)</h3><p>One of my favorite parts of the new movie was the clear criticism of the binary way of looking at things. Early in the movie, Neo is making a game called “Binary” which is going nowhere. While Binary was a character in the original trilogy, I don’t think this is the reference in this case. Throughout the movie, we keep seeing blurred lines. Nothing is as “binary” as it was in the original trilogy anymore. Agent Smith is now “woke” (sorry, had to). There are machines becoming sentient and are aligning with humans. Going back to the original point, this again is a different take than the original trilogy which had clear lines between good and evil, machines vs. humans. This is no longer the case. This blurring of lines is resulting in wonders in Io, which were not possible before. But at the same time, the same blurring is resulting in humans becoming complacent in their own simulation.</p><p>Possibly reading too much into it, I think this is a reference to how ubiquitous technology and “AI” has become in our lives. Elon Musk has argued that we are already cyborgs, which has some truth to it considering how dependent we are on our technology. Especially in the space of AI-driven applications, we are seeing incredible advances. While as a practitioner I know there are so many things that simply don’t work in the way these algorithms are built, as a user I can’t help but being impressed by their seamless integration into our devices. The integration of machines into human world (the “real” world) is a not so thinly-veiled reference to this phenomenon we are experiencing. As this is something that I am personally and professionally interested in, I was very pleased to see it represented.</p><h3>Potential vs. Actualization</h3><p>The original trilogy was rather anarchist. A significant portion of the adult viewers of the new movie were teenagers back then. As teenagers, we all found the message and the act of escaping from a reality that we didn’t want to be in very compelling. This isn’t to say the original trilogy was nothing but a means to appease our teenage angst. It was much more, if you watched it later as an adult, you would still be compelled to argue for or against the messages in the movie. This makes the original trilogy a truly captivating piece of work.</p><p>Except it was too utopic.</p><p>The original Matrix trilogy is based on the premise that if people were presented with the choice of knowing the truth in exchange for their comfort, they would choose the truth. When the option is presented, Neo barely hesitates and jumps right in. While he has some self-doubt during the entire trilogy (which is called character development), he doesn’t actually second-guess his choice. He just doesn’t believe he is the one. Not only everything is crystal clear (i.e. binary) but also, he, as a human has so much potential. Along with the rest of the “woke” humans, residents of Zion.</p><blockquote>The original trilogy is about human potential, <em>the promise of humans. </em>The new one is about why they will fail to achieve it.</blockquote><p>I thought this was an impressive take.</p><p>In the new movie, we see Neo doubting everything he has been told. His “therapist” keeps him drugged with not only actual drugs but comforting lies. Neo chooses the comforting lies over the unpleasant truth that he already knows for so long. Humans have become complacent. Trinity is straight up a soccer mom. “Blue to the core” both of them. It takes quite a bit to wake Neo up and when he does he is so drained that he can’t fly. Trinity almost doesn’t give up her sweet comfort.</p><p>Back in the “woke” people world, former fighters and residents of Zion are now in Io. Which is a hide-out city. They are no longer fighting. Instead, they are cowering behind a false curtain and trying to grow strawberries. The crew, which is basically extracted from Sense8, gets scolded and punished for rescuing Neo, the One, in the first place. Humans are settling for a comfortable-but-restricted life instead of complete freedom from the enslaving machines. They are no longer trying to liberate other humans.</p><blockquote><em>The residents of Io sacrificed the truth of other humans in favor of their own comfort.</em></blockquote><p>This is a powerful message (that is admittedly glossed over and downplayed but more on that later). People who know me well will have heard me say this on multiple occasions. Humans are incredible in the potential they have for greatness, whether it be good or evil. They are also extremely disappointing because only very few of them will strive to achieve it. I loved seeing a semblance of this in the new movie but I wish it had been done better.</p><h3>So What Sucked?</h3><p>The movie could have been absolutely extraordinary if it weren’t entirely made to make money. The take on binary vs. non-binary worlds, and the decline of human society are really powerful topics that have been glossed over. Even the mocking of the original trilogy could have been done much better, and that was one of the things people hated about it (although I found it amusing). Instead, it was all glossed over in favor of the main plot.</p><p>Also the main plot sucked.</p><p>The whole long-lost-love and broken-hearted chosen one is getting really old. The fact that the salvation of Trinity was simply to show that she ends up being the chosen one is a really cheap trick that takes so much away from a movie that could have been great.</p><h3>Final Words</h3><p>In general, I think if you go into the movie by not expecting anything on the caliber of the original trilogy, but rather something that is based on it, anyone who is not hell-bent on hating the movie would find things to enjoy. Having said that, the new movie is targeted towards an audience who consumes shallow bits of non-informative signal all day every day, therefore its depth is very lacking. And the fans of the original trilogy are so deprived of anything of that depth, I believe they (as well as I) were hoping for something that would satisfy that yearning. Unfortunately, this was not it. Yet, instead of hating it and dismissing it outright, I advocate for looking at the glossed-over themes in the movie because honestly, what do you hope to achieve by hating it in the first place? Nobody gives a fuck if you think it is canon or not, it is now out there and in the books.</p><p>7/10</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e12962053d77" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Data-Driven Tool to Unwind FA Player Eligibility Saga]]></title>
            <link>https://medium.com/ensk-ai/a-data-driven-tool-to-unwind-fa-player-eligibility-saga-1f1a6ab76734?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/1f1a6ab76734</guid>
            <category><![CDATA[football]]></category>
            <category><![CDATA[sports]]></category>
            <category><![CDATA[soccer]]></category>
            <category><![CDATA[uk]]></category>
            <category><![CDATA[analytics]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Mon, 26 Apr 2021 09:08:53 GMT</pubDate>
            <atom:updated>2021-04-26T09:08:53.962Z</atom:updated>
            <content:encoded><![CDATA[<h4>FA’s rules for football player eligibility to play in UK is just as confusing as Brexit deal itself, but fear not, we have got you covered.</h4><p>In the aftermath of Brexit, the <a href="https://www.thefa.com/">British Football Association </a>announced a set of rules that govern the eligibility of a football player to play in the UK. And obviously, it had to be as convoluted as possible, otherwise it wouldn’t be on brand for the only European country that drives on the other side of the road, and <a href="https://www.youtube.com/watch?v=vDBzi0n9Fxg">still has a law about running around with a suspicious salmon</a> (Watch the video it is hilario<strong>u</strong>s).</p><p>In any case, our club and agent clients were having a hard time navigating the new landscape so we built a tool to remove the hassle. A tool we are now making public <strong>with no strings attached.<em> Because of course as Europeans we have to come to the rescue.</em></strong></p><blockquote>The tool allows you to search any name, and it will return the results if the player exists in our database of &gt;100k active football players.</blockquote><blockquote>Behold the TOOL: <a href="https://caniplayintheuk.com/">https://caniplayintheuk.com/</a></blockquote><p>Basically, all you need to do is type in the name of a player and poof, you have the verdict. Pretty magical.</p><h4><strong>Method to Our Magic</strong></h4><p>Even magic has a method to it so we should probably explain how it actually works. Simply put, we took the GBE rulebook and codified all the conditions which award eligibility points to players. We then run these rules on top of our data for all active players in the world and calculate the points obtained by each player based on their actual performance. Voila! The results are updated after every game and are always up to date.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SD3bG91COVj6b4WSlrj-wg.png" /><figcaption>Screenshot of the tool</figcaption></figure><p>Of course some rules are open to interpretation, so we also explain distribution of where the players get points from. As a general rule, we hate black-box approaches in EnskAI so we put a lot of effort in explainability and interpretability. In this case, the solution is not a complicated model but rather a calculator, yet, we still follow our core principles.</p><h4>The Part Where We Do Blatant Self-Promotion</h4><p>Of course this is not all we do with data, but it is a good example of our philosophy of having a global reach and a practical, transparent approach in our solutions. What this means is, we prefer being able to objectively analyse all data points available to us (for instance players or teams across the globe) in a fair and comparable manner. <em>Long story short, we are able to analyse every single team and every single player on the planet in conjunction based on solid understanding of football mechanics. </em>This helps our clients a lot when they want to find a player that fits their needs from an obscure competition and know <em>exactly</em> how he compares to any player they are familiar with. Same with the clubs when they want to know exactly how they compare to their opponents. How we do this is pretty cool, and if you want to know more just reach out. We are more than happy to walk you through the math behind the magic.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1f1a6ab76734" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ensk-ai/a-data-driven-tool-to-unwind-fa-player-eligibility-saga-1f1a6ab76734">A Data-Driven Tool to Unwind FA Player Eligibility Saga</a> was originally published in <a href="https://medium.com/ensk-ai">Ensk.AI</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[On The Wholesomeness of Climbing]]></title>
            <link>https://l1-aeg.medium.com/on-the-wholesomeness-of-climbing-befb0c3c07f3?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/befb0c3c07f3</guid>
            <category><![CDATA[mental-health]]></category>
            <category><![CDATA[sports]]></category>
            <category><![CDATA[climbing]]></category>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[mindfulness]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Wed, 31 Mar 2021 20:57:29 GMT</pubDate>
            <atom:updated>2021-03-31T20:57:29.264Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GXIVGhx_CvmaCqjqLthBVg.png" /><figcaption>Drawn: The Art of Ascent: <a href="https://www.amazon.com/Drawn-Art-Ascent-Jeremy-Collins/dp/1594859582">https://www.amazon.com/Drawn-Art-Ascent-Jeremy-Collins/dp/1594859582</a></figcaption></figure><h4>Climbing is about mindfully controlling your body, your mind, your emotions and your attitude; and letting go of anything else.</h4><p>I recently started climbing just because a couple of my friends were going climbing and I had nothing better to do. As someone who has a fear of heights, I basically just wanted to see whether I would be able to control my fear in a controlled setting. The experience turned out to be mentally and emotionally far more complex than I had anticipated. TLDR: I got hooked, but that had nothing to do with the adrenaline rush.</p><p>Before starting this personal article, I searched Medium a bit to see if my points were already made and found an article that reflects on the benefits of climbing from a professional perspective and I wholeheartedly agree. The article is <a href="https://medium.com/swlh/five-life-lessons-from-climbing-ddd35eadd73e">here</a>. My piece is going to be a lot more personal but still touching on similar aspects.</p><p>Past couple of years have been rather stressful due to work, whole lot of life changes and academic studies. I have been fortunate enough to be able to rely on close friends and family for support but in general, my personal experience has been that the only way out of that burnt-out and frankly depressed mental state is by working through it, bit by bit. Support of loved ones makes it easier to keep going, but you still have to do the work. Friendly disclaimer at this point, please do reach out for professional help if you feel like you need support. Please also reach out even if you don’t think you need it. Sometimes you just don’t know. Hopefully, professional mental help becomes more accessible and affordable to everyone and fast.</p><h4>Exercise As a (Personal) Mental Treatment</h4><p>During the early days of that state of mind, I started walking gradually longer distances because it helped clear my head. The amount of distance I covered reached 15 km on a working day. At this point it became unsustainable due to time constraints so I started running. While I have a history of doing various sports (gymnastics, skiing, volleyball, tennis, table tennis), some of them on a competitive level, it was a very long time ago. Basically I hadn’t done any real sports since university at the time and I was getting into it because I needed a release.</p><p>Running worked well, because once you are that tired, all you can think of is just putting one foot in front of the other. There is solace in that simplicity. Runner’s high is also very real for me, so it helps a lot with mental state.</p><p>Eventually my running evolved into outdoors and I started doing more and more challenging hikes along with running. Part of the challenge of the hikes, especially the longer and and more technical ones, is mental but on the other end of the spectrum. Some hikes put you in a precarious position and when at a precarious position, your mind gets to a hyper-drive. It is not exactly an adrenaline rush. If you can make it past that stage of initial panic, your awareness is <em>heightened</em>.</p><p>So at that point, I basically had two ways of working through my problems: running or hiking them off. Running helps because when I exhaust myself to the point where my main problems become more fundamental, like being able to breathe. With hiking, the experience is much more calmer, and during the majority of the hike you are actually able to think and work through the issues until you again get exhausted enough to start thinking about more basic needs. The heightened awareness only happens when the scenery is extraordinarily beautiful or you are physically challenged.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YTuBJWPmyGYqCf3W.jpg" /><figcaption>Source: <a href="https://www.spain.info/en/discover-spain/climbing-mountaineering/">https://www.spain.info/en/discover-spain/climbing-mountaineering/</a></figcaption></figure><h4>The Late Christmas Gift</h4><p>So when I climbed for the first time (December 28, 2020), it was everything I was looking for and more. Climbing, especially if you have 0 technique , is physically <em>very</em> demanding. First time I climbed, I obviously had no technique and my mind also didn’t acknowledge that I was tied to the end of a rope and couldn’t possibly fall. So basically, for me the exercise was exerting quite a lot of physical effort while at the same time being incredibly aware of where I am and <em>how</em> I am. Whether I am balanced or not (I was not), what is potentially my next move (they were all wrong), how far up I am from the ground (not very far), <em>how do I eventually get down </em>(because I <em>obviously</em> was not going to let go and let myself be descended, so basically I spent the whole afternoon walking up and down walls). Couple days after this experience I dislocated my ankle running, so had to stop doing anything for weeks.</p><p>I got back to it on Feb 9. Same story as above, but this time I was determined to be able to let go and actually <em>descend with rope.</em> First couple of climbs, I didn’t manage. Finally I did, when I landed on the ground, I was trembling because frankly I had been terrified to let go. What I am about to say is a bit of a cliché and cheesy, but this is one of the things I am most proud of in this whole journey. Letting go, both literally and figuratively, has always been hard for me. That was a good lesson. This makes me proud.</p><h4>The Reflection on The Wall</h4><p>After that I legitimately started climbing. I quickly realised (or thought I did) that I am at a disadvantage because I am very short at 155 cm (5&#39;1&quot;). This was pretty discouraging, other people could reach holds just by lifting their arms where I had to jump to reach them. In most cases, this wasn’t even enough because I don’t have the physical power to make the jump or hold afterwards. At this point, you either quit or you accept you are at a physical disadvantage and work through it. I had found something that helped me immensely mentally, <em>I wasn’t about to let go</em> (I know.). So I decided to swallow the disappointment, feelings of inadequacy, creeping self-image issues and work through the routes.</p><p>Turns out I was wrong about my height being a disadvantage. It is not an advantage either. <em>It is just what it is, especially in climbing. </em>The lack of reach forced me to try to figure out other ways to top the routes. At this point, I was introduced to the concept of <em>boulder problems.</em> Boulder problems are climbing sequences that require you to figure out a way of progressing through. It is apparently standard climbing lingo, which was very confusing initially. The way you work through a boulder problem by body movement is called a <em>beta</em>. Different types of bodies come up with different <em>beta</em>s to solve the boulder problems. While I may not be able to reach a certain hold, I fit into smaller openings on the wall. The fact that my limbs are shorter also means that when I reach, I am upsetting my balance less than someone with longer limbs. Some moves are easier for me compared to a taller person, and some are harder.</p><blockquote>While climbing, you are forced to figure out a <em>solution</em> that works for you as an individual, making the experience about you and only you.</blockquote><p>I find this very profound. It forces you to stop and think about the problem ahead, what you are capable of doing and what you are not and how to combine your resources to overcome that problem. It forces you to <em>stop and think</em>.</p><h4>The Virtue of S<em>topping</em></h4><p>In my first initial climbs after learning to fall down, I tried to climb routes by basically rushing and running through them. This meant brute-forcing some sequences, which didn’t really work very well. Then, mostly to overcome my perceived height issue, I started to train some technique. This became easier, partly because I was getting stronger, and partly because I was using better technique thus spending less energy. But I was still rushing through the routes. The beginner routes allow you to do that because there are enough holds to keep going.</p><p>As I got to the intermediate level routes, that basically became impossible because they were physically as well as problem-wise more challenging (hand-holds and foot-holds getting smaller, more separated or sequenced not as intuitively). I still tried to rush but that didn’t work. So I was forced to stop and be more aware of the holds available to me. Climbing smart yielded better results and faster progress than climbing with brute-force. I found myself more easily working through the routes where I thought I had to jump or leap to reach by simply shifting my body position or extending a leg in a different angle to just reach to the hold statically. A mindset shift and consequent physical shifts made topping the routes so much easier.</p><h4>The High and The Low</h4><p>Emotionally, topping a route you have been struggling with feels incredibly rewarding. Even making progress on the routes is very rewarding. But also, there is a lot of frustration that comes with trying over and over again and <em>failing</em>. And you just simply have to deal with that. You have to just accept it wasn’t going to happen that time and be willing to come back again and again not knowing whether you will succeed eventually. This teaches resilience and humility. You can get frustrated all you want. The boulder doesn’t care. You can either beat yourself up about it, or just work through it by analysing why you are failing and keep improving physically and technically.</p><h4>The Control and The Acceptance</h4><p>Climbing overall is about <em>control</em>. Sometimes this control means <em>restraint</em> so you don’t rush and think instead. Sometimes the control means making that leap where you have to, however scary it may be. And sometimes it means swallowing a whole lot because you are not progressing in a route.</p><blockquote>Control your body. Control your mind. Control your emotions. Control your attitude. And anything else, let go.</blockquote><p>Yet another thing climbing forces you to accept is the limits of your body. I am not talking about the height per se, but physical capacity does impose a limitation on what you can do. And this can change from climbing session to session, even during a session. Climbing puts a lot of toll on your tendons. Your muscles will likely be fine, but your tendons, which are the tissues that tie your muscles don’t develop as fast. It is easy to hurt these structures. No matter how much you might want that send, it may simply not be possible because you lack any aspect of physical capacity needed. This can be endurance or power or strength or your finger simply may not have developed just yet. You are forced to be aware of your physical state, and <em>treat yourself kindly</em> if you eventually want to make the route. This is not very specific to climbing, but applies to all sports.</p><p>In fact, none of the above is climbing specific. But for me personally, climbing has been the only sport that combined all aspects of control that I seem to be needing. To be completely honest, writing a long <em>sonnet</em> to an activity that I have been doing for 7 weeks and climbing only at the level of 6a makes me feel a bit phony. But climbing did wonders for me mentally, possibly physically too, so I can’t help but share. Different things will work for different people. Climbing may not work for you. But as someone with a very loud and busy mind, and not very kind to herself, having found something that forces me to exercise this mindful control has been a true blessing. Hopefully, everyone can find something similar. We all deserve clarity.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=befb0c3c07f3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Comprehensive Guide to Boosting Models]]></title>
            <link>https://medium.com/data-science/comprehensive-guide-to-boosting-models-37714471e94d?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/37714471e94d</guid>
            <category><![CDATA[xgboost]]></category>
            <category><![CDATA[lightgbm]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[gradient-boosting]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Sat, 27 Mar 2021 09:15:08 GMT</pubDate>
            <atom:updated>2021-03-31T19:29:22.894Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*A_dQOEwDz9ug_sGS" /><figcaption>Image Source: <a href="https://twitter.com/wrathofgnon/status/1250287741247426565/photo/1">https://twitter.com/wrathofgnon/status/1250287741247426565/photo/1</a></figcaption></figure><h4>In the world of modelling tabular data, there is one ruling dynasty: Boosting models.</h4><p>While the landscape is changing, majority of the problems in ML and AI continue to be supervised modelling problems. Despite deep learning applications (image, video, text, speech) taking the headlines left and right, majority of the real life problems use tabular data, especially in business. And in this world there is only one ruler: Boosting models.</p><p>People who have participated in Kaggle competitions are very familiar with boosting models. The reign started with XGBoost dominating most competitions. The crown was shortly passed onto LightGBM. But these are not the only boosting algorithms. Recently, I have been asked a question on AdaBoost which made me realise I had overlooked this bit of boosting legacy. So I decided to write a comprehensive guide, mostly as a self-reference but also as a summary for everyone else. This blog-post aims to discuss the structural differences between these models and their performance.</p><h3>A Primer on Ensembling and Boosting</h3><p>Before we dive into different models, let’s quickly take a look into what boosting is and its cousins. Boosting is a method for ensembling outputs, which sounds rather cryptic if you haven’t been introduced to the concept. Ensembling is a way to combine outputs from different models/decisions/functions. Boosting is one way of ensembling. There are three main methods of ensembling: Bagging, boosting and stacking.</p><p>Basically in ensembling, we have a bunch of simpler models (so-called ‘weak-learners’), which ideally capture different aspects of our data, and we are using all of their judgement to arrive at an optimal solution. Sort of like algorithmic representative democracy. And just like in democracies, if the opinions are very similar you end up in an echo chamber that overfits to one aspect of your data and your model will fail in practice. So our ‘weak-learners’ ideally have low-correlation. For a more-formal and less-opinionated explanation of this, see bias-variance trade-off.</p><p>In boosting/bagging <em>tree </em>algorithms, when we say weak-learner we actually mean decision trees. But weak-learner is a generic term used for any ML model that performs slightly better than random chance.</p><h4>Bagging</h4><p>Bagging is the method where we use the decisions coming from all our weak-learners altogether. Basically, you have a vote coming from each of the weak-learners, you make the decision to use most common opinion (in case of classification) or the average (in case of regression). Assume we have 5 weak-learners, 3 of them say ‘yes’, 2 of them say ‘no’ to the question of ‘Would this person survive Titanic?’, the final decision would be ‘yes’. Random Forest is the most well-known bagging algorithm in this context.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/592/0*3bEG9qr6-Tdmcrdo.png" /><figcaption>Image Source: <a href="https://en.wikipedia.org/wiki/File:Random_forest_diagram_complete.png">https://en.wikipedia.org/wiki/File:Random_forest_diagram_complete.png</a></figcaption></figure><h4>Boosting</h4><p>On the other hand boosting approaches the problem differently. It iteratively builds its opinion by adjusting for errors at each step. Each weak-learner is built on-top of the previous set of learners. For instance, let’s say we already have 5 weak-learners built already, the combined output decision of these 5 learners are assumed fixed, the 6th learner is expected to correct their mistakes (errors) compared to the training output. In boosting, each weak-learner is weighted in one manner or another to optimise the loss function. This is why it is called ‘boosting’. It ‘boosts’ the performance of previous set of models. This weighting mechanism is one way that boosting algorithms differ from each other.</p><h4>Stacking</h4><p>So to recap, bagging averages the outputs. Boosting combines the outputs with some weight. In that sense (apart from conjoint and iterative modelling approach) bagging and boosting differ in their weighting approach. Bagging has equal weights for all weak learners, boosting weighs the weak-learners based on their performance on loss function. In contrast, <em>stacking</em> uses <em>yet another</em> learner to learn how to combine the weak learners. The stacking layer can have any model as a combinator of all the weak learners. So basically instead of weights, you can use, say, neural networks to combine the outputs coming from your weak learners.</p><h3>Boosting Models: AKA How do we create ‘John Watson’ from ‘David Bowie’s?</h3><p>If you have read any of my previous writings, you would know I can never resist an obscure quote, but this one is so fitting and not really all that obscure, I have to give a <a href="https://www.goodreads.com/quotes/500086-if-you-took-a-couple-of-david-bowies-and-stuck#:~:text=Quotes%20%3E%20Quotable%20Quote-,%E2%80%9CIf%20you%20took%20a%20couple%20of%20David%20Bowies%20and%20stuck,have%20something%20which%20didn&#39;t">reference</a>. Now that I have explained what ensembling does in the <em>clearest way possible</em> by means of this quote, I can move on to listing different boosting models below.</p><ul><li>AdaBoost</li><li>Gradient Boosting Machines</li><li>XGBoost</li><li>CatBoost</li><li>LightGBM</li></ul><h4>AdaBoost</h4><p>AdaBoost is the original boosting model that stands for Adaptive Boosting. At each boosting iteration, it adjusts the weights based on model’s current performance on each of the samples. Using these weights, the dataset is re-sampled (with replacement) into a new training set. Basically by increasing the weight on each misclassified sample and reducing the weight on correctly classified ones, we are increasing the chances these misclassified samples will appear in our training subset (maybe more than once) in the next iteration. Therefore, the model continuously re-focuses on items we are not able to classify to find some way of representing them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/0*4asIMx1EB4k47UzO.gif" /><figcaption>AdaBoost in Action — Source: <a href="http://srl.informatik.uni-freiburg.de/videos">http://srl.informatik.uni-freiburg.de/videos</a></figcaption></figure><h4>Gradient Boosting Machines</h4><p>As previously mentioned, not all boosting models do updates like AdaBoost. In GBM’s case, the model updates are done via gradient descent given a loss function as opposed to changing sampling weights of the data. The update rule is as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/908/1*xLg106nlHyQcqSJ6Kixz9A.png" /><figcaption>Image source: <a href="https://en.wikipedia.org/wiki/Gradient_boosting">Wikipedia</a></figcaption></figure><p>Basically, what this is telling you is, at iteration <em>m</em>, select the weak-learner <em>h</em> that minimises the loss between actual value <em>y_i</em> and our previous estimation <em>F_{m-1}(x_i) + </em>prediction of<em> h </em>(i.e. the weak learner). The way to find the <em>h</em> that minimises this is through gradient descent.</p><p>In case of optimising mean squared error for regression, the gradient is simply the residuals. And each consecutive weak learner is fit to the <em>residuals</em> because we are trying to compensate for errors. Imagine this scenario, when we start, we have 0 clue about our outputs. Therefore, every value in our output vector <em>y</em> is actually a residual, error. So we start by fitting a weak learner to our initial residuals. Then, our weak learner makes some predictions, some of them are better than others, so we try to learn the errors and keep going as such. This process is equivalent to training boosting models using gradient descent. Math is fairly simple but too long to put here. For classification and other regression problems that do not optimise MSE, we train our weak learners in such a way that they will perform gradient descent. GBMs are therefore generalised boosting methods that can fit to any differentiable loss function. <a href="https://www.ccs.neu.edu/home/vip/teach/MLcourse/4_boosting/slides/gradient_boosting.pdf">Here</a> is a good set of lecture notes on the topic which also shows the equivalence of fitting to residuals and gradient descent.</p><h4>eXtreme Gradient Boosting (AKA XGBoost)</h4><p>XGBoost is a regularised variant of GBMs. Instead of minimising the loss as above in GBM, XGBoost adds a term for regularisation to the loss function. This is a simple but very effective trick that prevents overfitting. This regularisation term enforces a minimum level of required gradient for the updates to apply. If the regularisation term is much larger than the gradients, then the gradients will have no effect at all, so the updates are stopped. If the regularisation term is infinitesimally small, then this update is equivalent to GBM. So instead of minimising the <em>argmin</em> part of GBM equation above, we are now finding the learner <em>f_t </em>that minimises the following:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/722/1*sjRgPJLqLZLYk1j75OGvHg.png" /><figcaption>XGBoost learning. Image Source: <a href="https://xgboost.readthedocs.io/en/latest/tutorials/model.html#">XGBoost docs</a>.</figcaption></figure><p>In addition, XGBoost adds proportional shrinkage of leaf nodes to further prevent overfitting and additional processor performance related changes. X in XGBoost has nothing to do with data science or ML related concepts, it points more towards the engineering focus developers had to bring a distributed and fast algorithm to the ML space.</p><h4>CatBoost</h4><p>While a great algorithm for tabular data, XGBoost does not handle categorical values natively. They must be manually handled by some form of numerical encoding. CatBoost on the other hand, is able to handle categorical variables as long as you specify the column names. The way this works is, it generates dummy (one-hot encoded) variables up to the max number of columns provided by <em>one_hot_max_size </em>parameter. If categorical variable has more unique values than this parameter, a clever permutation and subsequent class averaging is used to create an average target value per class and category. The formula is given below, it looks slightly complicated but here is what it means: For a given permutation, find count the number of samples up to this point (j) where <em>y (output variable)</em> is equal to 1 (assuming binary classification), add some prior to it and divide by total number of objects up to this point.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8kWvJlyMPCKta8c6.png" /></figure><p>That means a couple of things, given a feature <em>x</em>, we permute it <em>n</em> times to obtain different orderings of this feature, then sample by sample, we calculate the ratio of positive class with some offset to the number of entries so far. That means, in the first instance, if we have class value <em>a</em>, and <em>y=1</em> that means our replacement value is (1+ a.P)/(1+a), in the second instance let’s say we have class value <em>b</em> and <em>y = 0</em>, then our replacement would be (0 + a.P)/(2 + a). As such we keep going. Notice that we are scanning through all samples using an extending window so this has the potential to get very slow with high number of examples. Essentially, this formula is the <em>probability of observing a positive class for this value of the category in this feature</em> using a Bayesian online estimate. The “Bayesian” component of it comes from the prior (P) and coefficient <em>a </em>is how much we trust the initial prior probability. To be fair this is not really an original idea, it has been used in categorical representation by many data scientists. CatBoost was the first boosting ML model that formalised it. It works well for categorical features with large number of categories however DS practitioners can implement this without having to rely on CatBoost directly.</p><h4>LightGBM</h4><p>LightGBM is yet another improvement on XGBoost, it can also work with categorical values directly. This model implements a couple of structural changes.</p><ul><li>Different sampling strategy while computing splits: Gradient-based one side sampling (GOSS). XGBoost uses a histogram based splitting strategy. Basically it computes a histogram of the data, and performs tree splits based on the <em>information gain</em> on histogram. On the other hand, LightGBM performs gradient based sampling first and then computes the split based on the sample subset. The way this works is, it retains all the rows with high gradients and samples x% of the rows with low gradients, then builds a new model on this subset. The pseudocode of this algorithm can be found in the paper: <a href="https://papers.nips.cc/paper/2017/file/6449f44a102fde848669bdd9eb6b76fa-Paper.pdf">https://papers.nips.cc/paper/2017/file/6449f44a102fde848669bdd9eb6b76fa-Paper.pdf</a></li><li>Efficient feature-bundling to reduce the number of features via pruning of exclusive feature sets. The idea is, you can group features that have no overlap between their non-zero values into one feature.</li><li>Leaf-wise splits vs level wise splits: LightGBM grows the leaf that provides maximum reduction in the loss function. Theoretically, the tree can grow on one leaf-side only. Other algorithms by default grow the tree level-wise, meaning tree only grows one level at a time. XGBoost can also perform leaf-wise growth. This increases the speed and the efficiency of the algorithm, as it is a greedy search mechanism, however can lead to overfitting, especially if the data is sparse and/or small.</li></ul><h3>Conclusion and Practical Resources</h3><p>This post simply collects the design differences in algorithms however based on these design decisions, there are performance differences amongst algorithms.</p><ul><li>GBM is the OG (original gradient), performs well with different datasets however it cannot handle multiple real-life scenarios such as NULLs or categorical values. It has implementations in multiple standard libraries such as scikit-learn, so always a good idea to use as one of the baseline models. I try to use multiple ones: GLMs, decision trees, Random Forest and GBM to see which level of complexity gives a baseline acceptable performance. It guides the following decisions on model selection but I digress for now.</li><li>XGBoost is a significant performance improvement over GBM, but is not implemented natively in sklearn. Also cannot handle categorical variables. Can handle NULLs.</li><li>CatBoost is an improvement over XGBoost for categorical value increase, but the training efficiency is low. Works well with large number of unique feature values. Not implemented in sklearn natively.</li><li>LightGBM is an improvement over XGBoost that can handle categorical values and it is also very efficient however it is prune to overfitting unless carefully tuned. Not implemented in sklearn natively. However, the Python package has a sklearn submodule that is compliant with sklearn fit-transform syntax and it integrates with sklearn very well.</li></ul><p>For practical tips on how to tune and use these algorithms, below are a set of blog-posts I have personally referred to and continue to refer to. Official documentations of these packages are also very good so make sure you refer to those. Finally, it is a good idea to check out the original papers as well because they provide a way to read about the ideas of the creators in a more structured way.</p><ul><li><a href="https://towardsdatascience.com/lightgbm-vs-xgboost-which-algorithm-win-the-race-1ff7dd4917d">XGB vs. LGBM</a> Parameter tuning</li><li><a href="https://towardsdatascience.com/catboost-vs-light-gbm-vs-xgboost-5f93620723db">CatBoost vs. XGB vs. LGBM parameter comparisons</a></li><li><a href="https://xgboost.readthedocs.io/en/latest/">XGBoost documentation</a></li><li><a href="https://catboost.ai/docs/concepts/about.html">CatBoost documentation</a></li><li><a href="https://arxiv.org/pdf/1706.09516.pdf">CatBoost Paper</a></li><li><a href="http://learningsys.org/nips17/assets/papers/paper_11.pdf">CatBoost Other Paper</a></li><li><a href="https://lightgbm.readthedocs.io/en/latest/">LightGBM Documentation</a></li><li><a href="https://papers.nips.cc/paper/2017/file/6449f44a102fde848669bdd9eb6b76fa-Paper.pdf">LightGBM Paper</a></li><li><a href="https://github.com/microsoft/LightGBM">LightGBM Repo</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=37714471e94d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/comprehensive-guide-to-boosting-models-37714471e94d">Comprehensive Guide to Boosting Models</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Quantifying the Value of Indulgences]]></title>
            <link>https://l1-aeg.medium.com/quantifying-the-value-of-indulgences-d3bf4c92075?source=rss-c8c90cb6e013------2</link>
            <guid isPermaLink="false">https://medium.com/p/d3bf4c92075</guid>
            <category><![CDATA[values]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[habits]]></category>
            <category><![CDATA[social-media]]></category>
            <dc:creator><![CDATA[Elvan Aydemir]]></dc:creator>
            <pubDate>Sat, 20 Feb 2021 21:52:48 GMT</pubDate>
            <atom:updated>2021-02-20T21:52:48.219Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/0*2oqXULp-NHG8WSnS" /></figure><p>Sometimes, indulgences are ways to blow off steam, other times they are detrimental. A simple “utilitarian” principle might help us separate one from the other.</p><p>We all have indulgences and self-indulgences. In moderation, these indulgences help us feel better or push through some moments that may otherwise be overwhelming (looking at you chocolate cake). But at some point, if they turn into patterns we fall back to, they have a habit of becoming harmful, even turning into addictions.</p><p>Being human, we are never going to be completely efficient and calculated. To be honest, I think we absolutely mustn’t. Interesting and wonderful things happen when people make happy mistakes that open up whole new ways of doing things. That isn’t to say we should be completely chaotic either.</p><p>There is a fine balance between planning and experiencing.</p><p>When it comes to indulgences, I would like to keep the ones that improve my general well-being and remove (or reduce) the indulgences that are detrimental. Obviously, what is good and what is not is different for everyone. Personally, I wanted to pseudo-quantify the utility I got from all my indulgences.</p><p>Now is probably a good time to define what I mean by indulgence. I consider everything that I either enjoy or spend time on, that is not intended to contribute to my development as a human being (or is fundamentally a part of life) to be an indulgence. For instance, ice cream is an indulgence, a book is a necessity.</p><p>I can group my indulgences in 3 big buckets: Food &amp; Drinks, Shopping and Social Media. As I was trying to reduce my indulgences, I came up with a simple rule:</p><blockquote>If the utility I get from that action justifies the cost of doing it at the time, I do it.</blockquote><h3>Snacks &amp; Drinks: Battling Gluttony</h3><p>I love snacks. I simply love nibbling on food. But when you love snacks as much as I do, it is mostly the case where you just eat something and it is just… underwhelming. So, in basic utilitarian fashion I outlined above, I decided to evaluate all consumption with the following way of thinking: Am I going to enjoy this food enough to justify the calories I am consuming from it?</p><p>Keep in mind, this is about snacks and not about nutrition, I keep a healthy well rounded diet and consume enough calories. This refers to anything <em>on top</em> of my daily nutrition. An indulgence.</p><p>The answer to this question changes from snack to snack, or even for the time of the day for the exact same snack. In most cases, my answer is no, because most stuff is underwhelming. But sometimes, you just need that glass of wine or that piece of chocolate. So, then the condition is satisfied. Every additional unit of snack brings diminishing returns so there is a tipping point. Once I feel the scales tip I stop. I have been practicing this for about 8 months now, and I don’t have guilt or regrets about having a snack anymore. Worked wonders.</p><p>Same philosophy compelled me to stop eating meat. I love a good steak, or a beef carpaccio. However, there came a point in my life where I could no longer justify the personal cost (i.e. guilt mostly) attached to eating meat. This was 6 months ago, give or take. Everyone will have a different definition of utility and cost, and that is ok. I think this can still help though.</p><h3>Shopping: Battling Greed</h3><p>Another thing I love is shiny new toys. Mostly these are clothes, shoes and electronics. As I was moving apartments, I cleared my wardrobe and donated a lot of the stuff that was basically brand new because I had never worn them. I also have a tendency to buy the newest gadgets. So during the apartment change, I decided that I will not buy anything that I am absolutely not going to use.</p><p>This is arguably the most quantifiable case of indulgence, it literally comes with a price tag. The utility definition still is very person dependent. For instance, I recently replaced my vacuum cleaner because the old one was basically dead. The new one I bought would have been excessive in most circumstances, however, given how bad I am at cleaning and the fact that I have 3 cats, I decided to make the investment. On the flip side, I did not buy so many things, because I knew I was not going to use it therefore get no benefit from it.</p><h3>Social Media: Battling Vanity</h3><p>This revelation came very recently. I use social media quite a bit. And to be honest, I get a lot out of it too, mostly in form of memes and interesting takes (good or bad). However, it has come to a point where I started feeling like different platforms provided different levels of utility.</p><p>In this case, I measure the utility of a social media platform as the ability to facilitate meaningful human interactions. Arguably, social media platforms are not the best places to do that, even though all of them have some implicit claim to be designed to do just that. But I have experienced numerous cases where I had meaningful interactions with people I didn’t know over these platforms.</p><p>Based on this criterion, <em>Twitter</em> tops my list of social media platforms that facilitate meaningful interactions despite allowing for only short messages. If you can ignore the trolls and the bitter people, you can actually find a lot of people with very interesting ideas on Twitter and directly interact with them in a non-pretentious way.</p><p>On the other hand, despite being explicitly designed to facilitate meaningful <em>professional</em> interactions, LinkedIn leaves me with mixed feelings. While I get a lot out of it in terms of knowledge and opportunities, the interactions have a sense of pretence that takes a lot away from the experience. Still provides positive utility though.</p><p>Finally, <em>Instagram</em>. This is where the revelation happened. I get (or got) absolutely nothing out of this platform. There is no meaningful interaction on this platform whatsoever because it is not designed to facilitate interaction, but to facilitate <em>broadcasting</em>. I don’t have much to broadcast in terms of visuals, aside from cat pics and the occasional bad photography. It is also not what I am looking for in terms of utility, so I decided I have been using this platform incorrectly for my own purposes. For now, it didn’t pass my threshold of providing positive utility.</p><p>This is not meant as an advice or anything, it is mainly a brain-dump of a simple principle that I use to make better decisions on superficial things. For more important things, it is probably good idea to consider multiple angles than calculate a simple pseudo-utility but for most daily decisions, I think this is fairly applicable. Enjoy.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d3bf4c92075" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>