<?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 Daniel Tsinis on Medium]]></title>
        <description><![CDATA[Stories by Daniel Tsinis on Medium]]></description>
        <link>https://medium.com/@danieltsinis?source=rss-d4fd4e2ad1d3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*IGA4G2K5EvxUFVGvKZlRtQ.jpeg</url>
            <title>Stories by Daniel Tsinis on Medium</title>
            <link>https://medium.com/@danieltsinis?source=rss-d4fd4e2ad1d3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 10:46:07 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@danieltsinis/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[Reason for Heuristic Models]]></title>
            <link>https://medium.com/@danieltsinis/reasoning-for-heuristic-models-832dc8d29ab0?source=rss-d4fd4e2ad1d3------2</link>
            <guid isPermaLink="false">https://medium.com/p/832dc8d29ab0</guid>
            <category><![CDATA[statistics]]></category>
            <category><![CDATA[articles]]></category>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Daniel Tsinis]]></dc:creator>
            <pubDate>Thu, 12 Jul 2018 00:01:13 GMT</pubDate>
            <atom:updated>2018-07-12T00:31:17.076Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*CK_4dgldd6Q8Hlk6TDD7_Q.png" /></figure><p>This article will aim to highlight an important idea: the basis of well performing models does not necessarily require them to be overly complex. In our current era of data modeling, we are often occupied with the creation of machine learning models that will be utilized as decision support tools; these models often come with lots of overhead. I have learned that when applied in a real world setting, simple models, which are heuristic in nature, can be both computed mentally and be comparable to more sophisticated techniques.</p><p>Simple heuristic models, which are often a string of effortless actions, can have the upper edge over complex models as they are cheaper to produce, and have comparable performance in specific situations. I came across this idea on on April 20th, when I attended a lecture at Columbia University where NYU Professor <a href="http://cusp.nyu.edu/profiles/ravi-shroff/">Ravi Shroff</a> spoke about some of his research around this topic. Shroff is focused on solving urban issues and works with many NYC organizations to try and tackle various problems. This discussion was focused on <em>‘simple rules for decision making</em>’ and it was meant to show that simple heuristic models can perform relatively well in comparison to more complex models in certain situations. The decision-makers, in the examples provided, were police officers and judges, however, Shroff also went through a brief example of how doctors can utilize simple heuristic models. The talk begins by stating that ‘experts’ often use intuition and experience to judge certain outcomes. This can be potentially problematic because often times these decisions have strong ethical implications as they can affect the trajectory of another person’s life. Here, a judge makes a decision in a courtroom but is affected by extraneous variables, meaning factors which should not influence decisions in a courtroom setting (link to the paper:<a href="http://www.pnas.org/content/108/17/6889"> http://www.pnas.org/content/108/17/6889</a>).</p><p>This is a notable quote from the abstract of the paper:</p><blockquote>We test the common caricature of realism that justice is “what the judge ate for breakfast” in sequential parole decisions made by experienced judges. We record the judges’ two daily food breaks, which result in segmenting the deliberations of the day into three distinct “decision sessions.” We find that the percentage of favorable rulings drops gradually from ≈65% to nearly zero within each decision session and returns abruptly to ≈65% after a break.</blockquote><p>Shroff states that as opposed to having people make decisions on intuition and experience we should rely on models to help us make our decisions, which generally outperform intuition and experience. The first topic which Shroff spoke about was Stop and Frisk. Stop and Frisk is a practice implemented by the NYPD to allow an officer to temporarily detain someone given suspicion that they have a weapon on them. In New York Stop and Frisk is now ruled unconstitutional so conspicuously the amount of occurrences has dropped. However, often times the person which is stopped has no weapon on them and the reason for which the stop occurs is one of personal bias. The question posed whether we can use statistics/machine learning to make fewer unproductive stops. Looking at data from 2008–2012 there were 750,000 stops (publicly available on NYPD) of which only 3.5% had weapon recovery. Shroff showed that he was able to build a more robust model that would produce acceptable results. He expected that he would be able to use the complicated model in order to predict who was carrying a weapon, but he wanted to know whether or not he can use a simpler model that is not as computationally intensive and ultimately lead to a series of questions the police officer can mentally compute. He boiled down to having 3 features. He compared the performance of this 3 feature model by regressing on the features and rounding the resulting coefficients to obtain integer weights, this yielded 8% of stops recovering 50% of the weapons. For something that can be done through some quick questioning, this produced good results.</p><p>The creation of simple models can provide a new framework for coming up with solutions for problems. Just as washing hands used to be a foreign event in medical settings, and when widely adopted, this simple adjustment produced drastically positive results. Straightforward and easy to execute actions strung together may be the answer to your problem.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=832dc8d29ab0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Looking for them before they’re gone]]></title>
            <link>https://medium.com/@danieltsinis/looking-for-them-before-theyre-gone-f4ddc1cb33c?source=rss-d4fd4e2ad1d3------2</link>
            <guid isPermaLink="false">https://medium.com/p/f4ddc1cb33c</guid>
            <category><![CDATA[climate-change]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[national-parks]]></category>
            <category><![CDATA[environment]]></category>
            <dc:creator><![CDATA[Daniel Tsinis]]></dc:creator>
            <pubDate>Mon, 23 Jan 2017 16:55:59 GMT</pubDate>
            <atom:updated>2017-01-23T23:04:25.063Z</atom:updated>
            <content:encoded><![CDATA[<h4>A National Park Biodiversity Analysis</h4><p>Last week, in my attempt to find an interesting dataset I discovered that the National Park Service uploaded ‘<a href="https://www.kaggle.com/nationalparkservice/park-biodiversity">Biodiversity in National Parks, 2016</a>’. Seeing that title I knew that I wanted to devote a few hours this weekend to data analysis. As this is a well-sized (approximately 120,000 items) dataset this is just preliminary exercise.</p><p>In the United States we have 58 National Parks that are governed by the federal government, making them federal land. However, our National Parks make up a small portion of the total land area that is under federal jurisdiction. It is in our greatest interest to keep <strong><em>all</em></strong><em> </em>our<em> </em>federal land under protection as it harbors a fragile collection of flora and fauna; a diversity that can only be preserved in its natural environment.</p><p>In this first analysis, there is a heat-map distribution of the number of species in National Parks broken down by state. For those interested in seeing a large variety of organisms the highest biodiversity is found in these 5 states:</p><ol><li>California (22,000)</li><li>Alaska (11,740)</li><li>North Carolina/Tennessee (6,623)</li><li>Nevada (7,092)</li><li>Washington (7,054)</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*G4MeYO9PgqlW7nOxstKJOQ.png" /><figcaption>Edited by <a href="https://medium.com/u/a32933812112">stella</a></figcaption></figure><p>While the work that the National Parks Services has done in cataloguing and providing public access to this information is commendable, poor resource management of federal land, such as commercial exploitation and unsustainable practices leads to general environmental harm. This means that even land that is protected or left untouched can be impacted quite literally downstream. From the data collection that National Park Service provided here is a distribution of National Parks with the greatest number of endangered species.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YYlebE2OEzfOxwqOzMzpPQ.png" /><figcaption>Edited by <a href="https://medium.com/u/a32933812112">stella</a></figcaption></figure><p>Delving deeper into the dataset I found that while there is a huge variety of species living throughout our National Park system, the West Coast and Hawaii hold special distinction for the number of endangered species. Hawaii’s dense micro-climes give the obvious explanation for their placement in the above chart, and California rounds out the top five parks with the greatest number of endangered species. The reasons for this are beyond the scope of this initial analysis, but I would appreciate hearing from any experts that can provide insight.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f4ddc1cb33c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A Guide To Hiking the Burroughs Range Loop in the Catskills]]></title>
            <link>https://medium.com/@danieltsinis/a-guide-to-hiking-the-burroughs-range-loop-in-the-catskills-963944f456b1?source=rss-d4fd4e2ad1d3------2</link>
            <guid isPermaLink="false">https://medium.com/p/963944f456b1</guid>
            <category><![CDATA[nyc]]></category>
            <category><![CDATA[hiking]]></category>
            <category><![CDATA[outdoors]]></category>
            <category><![CDATA[catskills]]></category>
            <category><![CDATA[backpacking]]></category>
            <dc:creator><![CDATA[Daniel Tsinis]]></dc:creator>
            <pubDate>Wed, 12 Oct 2016 02:13:27 GMT</pubDate>
            <atom:updated>2016-10-12T16:24:35.554Z</atom:updated>
            <content:encoded><![CDATA[<p>Utilizing the long weekend that Mr. Columbus provided us with in his conquest to capture the Americas, I figured I would take a trip to the Catskills region in upstate New York and catch some fall foliage while attempting the Burroughs Range Loop in one day. I originally planned to do the hike over two days, however the plan quickly morphed into a one day mission.</p><p>After a night of rain in Central Jersey I woke up at 4:30am doubting whether or not I should even make the trip up, thinking to myself, how much better can the weather be 2 hours North (completely different to my learning). I said screw it and decided to go either way, worst comes to worst it would be my first backpacking trip in the rain. I had packed all my backpacking essentials: tent, MSR, gas, water filter, food, etc, in order to have a safety net incase I was not successful in completing the loop in a day. With my gear on my back, it added about 20 lbs… fairly light.</p><p>I was out of my house at about 5am, and arrived at Woodland Valley Parking at about 7:30am. For all of those planning on going into that area, the parking is the first parking lot on the right towards the end of the road. The loop which I was planning on doing went in the order of Mount Wittenberg first, then Cornell, and then lastly Slide. If you begin on the trail head which is right in front of the parking area you will begin with Slide (I do not recommend this track).<strong> In order to hit the peaks in my order you must cross the road and head to campsite 46. </strong>By this campsite you will see a small boulder and to the left of it you will see the red trail marker and a foot bridge. Cross the footbridge to get on the trail. Under the footbridge was one of the only two times I saw water over the entire trail. So if you are doing this over two days I would recommend to carry all of the water needed. After about 1 minute of walking the trail turned pretty vertical and changes from steep to heavily inclined several times.</p><p>When you start approaching Mt. Wittenberg you will definitely know it. While I was doing this solo, I do not recommend for anyone to do this alone. You will start to see large rock buildups 10–20 ft high which you will have to climb up with all your stuff on your back. It’s not difficult if you have good balance and some upper body strength. There are about 3–5 climbs like this and then you make it to the top. The top is beautiful so take it in, especially since all of the other peaks do not compare to this one.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5gSW1OrBN27zqwD09i2Blg.jpeg" /><figcaption>Top of Mt.Whittenberg</figcaption></figure><p>After about another hour of cruising you will reach the “top” of Cornell. It feels like an extension of Wittenberg, and is simply a boulder which you can stand on to get a scenic panoramic view.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*un_v9wN7dtm2s64GL261Eg.jpeg" /><figcaption>Top of Cornell</figcaption></figure><p>The next hour after that is downhill walking. Through the downhill walk you will see several signs for primitive campsites. The signs are yellow arrows generally pointing in the according direction of a flat area. Many people do overnight stops here as you have exercised a great amount up till this point. After the hour passes you will see a mountain in front of you, that is what you are going to ascend, so get ready for more vertical walking. The ascend takes about an hour and the view isn’t all that spectacular. There is some flat ground available to rest and eat. The following hour and a half is downhill walking. The rocks on the way down are unpleasant to walk on as they are unevenly paced. Once reaching the bottom you will reach a junction, turn right, you have .7 miles left to Slide Mountain Parking lot.</p><p>More downhill.</p><p>Then you will come down to the Slide Mt. parking lot. Turn right onto the road and then the next part of this loop is walking on the road till Giant Ledge Parking. Once you get to Giant Ledge right at the beginning you will see a sign for Wood Valley Parking 3.5 miles away. There’s about an hour long hike up till you see another intersection, the right is toward Wood Valley. From this point on it is another hour and a half of downhill walking, except for one steep ascend section, and then you are finished!</p><p>I began this loop at 8am and finished at 4:15pm. If you are looking just for views I would recommend ascending toward Mt. Wittenberg, resting and eating up there, and then returning back down as opposed to doing the whole thing loop. However, if you are spreading this out over a one or two night adventure then it can definitely be a pleasant walk!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=963944f456b1" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>