<?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 Gunjan Agicha on Medium]]></title>
        <description><![CDATA[Stories by Gunjan Agicha on Medium]]></description>
        <link>https://medium.com/@gunjanagicha?source=rss-aeb3f45c6182------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*FpwhFwG6fvGaUyDdazExrw.jpeg</url>
            <title>Stories by Gunjan Agicha on Medium</title>
            <link>https://medium.com/@gunjanagicha?source=rss-aeb3f45c6182------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 16:04:26 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@gunjanagicha/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[Autodesk Experience]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">I did a 3-month summer internship in Autodesk, after which I was extended a full-time return offer and also a part-time offer for 2&#x2026;</p><p class="medium-feed-link"><a href="https://gunjanagicha.medium.com/autodesk-experience-6b4cf9a71dca?source=rss-aeb3f45c6182------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://gunjanagicha.medium.com/autodesk-experience-6b4cf9a71dca?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/6b4cf9a71dca</guid>
            <category><![CDATA[project-management]]></category>
            <category><![CDATA[focus]]></category>
            <category><![CDATA[autodesk]]></category>
            <category><![CDATA[internships]]></category>
            <category><![CDATA[work]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Thu, 04 Mar 2021 00:25:31 GMT</pubDate>
            <atom:updated>2021-03-04T00:25:31.835Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Buying a house with a loan vs all cash]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.datadriveninvestor.com/buying-a-house-with-a-loan-vs-all-cash-eb511abdf122?source=rss-aeb3f45c6182------2"><img src="https://cdn-images-1.medium.com/max/848/1*2kz09V6gJ13fsoQur6vjVg.png" width="848"></a></p><p class="medium-feed-snippet">I am planning to buy an apartment for my parents, and so this thought naturally occurred to me. Should I pay up-front for the house or&#x2026;</p><p class="medium-feed-link"><a href="https://medium.datadriveninvestor.com/buying-a-house-with-a-loan-vs-all-cash-eb511abdf122?source=rss-aeb3f45c6182------2">Continue reading on DataDrivenInvestor »</a></p></div>]]></description>
            <link>https://medium.datadriveninvestor.com/buying-a-house-with-a-loan-vs-all-cash-eb511abdf122?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/eb511abdf122</guid>
            <category><![CDATA[finance]]></category>
            <category><![CDATA[mortgage]]></category>
            <category><![CDATA[money]]></category>
            <category><![CDATA[buying-a-home]]></category>
            <category><![CDATA[loans]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Mon, 22 Feb 2021 09:11:09 GMT</pubDate>
            <atom:updated>2021-02-23T13:38:00.024Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[How to invest your money]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://gunjanagicha.medium.com/how-to-invest-your-money-cbc627a8e094?source=rss-aeb3f45c6182------2"><img src="https://cdn-images-1.medium.com/max/1517/1*oYA7vClrxkzNu6FuIwsYMw.png" width="1517"></a></p><p class="medium-feed-snippet">At this point of time, I am in 20k debt and don&#x2019;t have any money to invest. So, I am studying all the ways to invest, so that in the next&#x2026;</p><p class="medium-feed-link"><a href="https://gunjanagicha.medium.com/how-to-invest-your-money-cbc627a8e094?source=rss-aeb3f45c6182------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://gunjanagicha.medium.com/how-to-invest-your-money-cbc627a8e094?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/cbc627a8e094</guid>
            <category><![CDATA[money-management]]></category>
            <category><![CDATA[money]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[financial-planning]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Thu, 28 Jan 2021 19:05:08 GMT</pubDate>
            <atom:updated>2021-02-17T04:21:52.624Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Internship Interview Experience]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">I started my Master&apos;s program in Computer Science at the University of Texas at Dallas in fall 2019. I am applying for Data Science&#x2026;</p><p class="medium-feed-link"><a href="https://gunjanagicha.medium.com/internship-interview-experience-ab6b8e95e790?source=rss-aeb3f45c6182------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://gunjanagicha.medium.com/internship-interview-experience-ab6b8e95e790?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/ab6b8e95e790</guid>
            <category><![CDATA[interview]]></category>
            <category><![CDATA[internships]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Thu, 12 Dec 2019 20:42:51 GMT</pubDate>
            <atom:updated>2021-02-17T04:18:43.388Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Regression Analysis]]></title>
            <link>https://gunjanagicha.medium.com/regression-analysis-3417d077a3d3?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/3417d077a3d3</guid>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Wed, 16 Oct 2019 17:45:35 GMT</pubDate>
            <atom:updated>2019-10-16T17:45:35.744Z</atom:updated>
            <content:encoded><![CDATA[<p>types of regression:</p><ul><li>Linear Regression</li><li>Logistic Regression</li><li>Polynomial Regression</li><li>Stepwise Regression</li><li>Ridge Regression</li><li>Lasso Regression</li><li>ElasticNet Regression</li></ul><p>Regression analysis is a form of predictive modelling technique which investigates the relationship between a <strong>dependent </strong>(target) and <strong>independent variable (s)</strong> (predictor). This technique is used for forecasting, time series modelling and finding the <a href="https://www.analyticsvidhya.com/blog/2015/06/establish-causality-events/">causal effect relationship</a> between the variables. For example, relationship between rash driving and number of road accidents by a driver is best studied through regression.</p><h3><strong>1. Linear Regression</strong></h3><p>In this technique, the dependent variable is continuous, independent variable(s) can be <a href="https://en.wikipedia.org/wiki/Continuous_and_discrete_variables">continuous or discrete</a>, and nature of regression line is linear.</p><p>Linear Regression establishes a relationship between <strong>dependent variable (Y)</strong> and one or more <strong>independent variables (X)</strong> using a <strong>best fit straight line</strong> (also known as regression line).</p><p>It is represented by an equation <strong>Y=a+b*X + e</strong>, where a is intercept, b is slope of the line and e is error term. This equation can be used to predict the value of target variable based on given predictor variable(s).</p><p>The difference between simple linear regression and multiple linear regression is that, multiple linear regression has (&gt;1) independent variables, whereas simple linear regression has only 1 independent variable.</p><h4>How to obtain best fit line (Value of a and b)?</h4><p>This task can be easily accomplished by Least Square Method. It is the most common method used for fitting a regression line. It calculates the best-fit line for the observed data by minimizing the sum of the squares of the vertical deviations from each data point to the line. Because the deviations are first squared, when added, there is no cancelling out between positive and negative values.</p><p>Model is evaluated using r-square metric</p><p>multiple regression suffers from multicollinearity.</p><p>its very sensitive to outliers.</p><p>to find significant variables we use:</p><p>a. forward selection</p><p>b. backward elimination</p><p>c. step wise approach</p><h3>2. Logistic Regression</h3><p>Logistic regression is used to find the probability of event=Success and event=Failure. We should use logistic regression when the dependent variable is binary (0/ 1, True/ False, Yes/ No) in nature. Here the value of Y ranges from 0 to 1 and it can represented by following equation.</p><pre>odds= p/ (1-p) = probability of event occurrence / probability of not event occurrence<br>ln(odds) = ln(p/(1-p))<br>logit(p) = ln(p/(1-p)) = b0+b1X1+b2X2+b3X3....+bkXk</pre><p>Above, p is the probability of presence of the characteristic of interest. A question that you should ask here is “why have we used log in the equation?”.</p><p>Since we are working here with a binomial distribution (dependent variable), we need to choose a link function which is best suited for this distribution. And, it is <a href="https://en.wikipedia.org/wiki/Logistic_function"><strong>logit</strong></a> function. In the equation above, the parameters are chosen to maximize the likelihood of observing the sample values rather than minimizing the sum of squared errors (like in ordinary regression).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/359/0*p4KJ4fbASqjFo2sB" /></figure><h4>Important Points:</h4><ul><li>It is widely used for <strong>classification problems</strong></li><li>Logistic regression doesn’t require linear relationship between dependent and independent variables. It can handle various types of relationships because it applies a non-linear log transformation to the predicted odds ratio</li><li>To avoid over fitting and under fitting, we should include all significant variables. A good approach to ensure this practice is to use a step wise method to estimate the logistic regression</li><li>It requires <strong>large sample sizes</strong> because maximum likelihood estimates are less powerful at low sample sizes than ordinary least square</li><li>The independent variables should not be correlated with each other i.e. <strong>no multi collinearity</strong>. However, we have the options to include interaction effects of categorical variables in the analysis and in the model.</li><li>If the values of dependent variable is ordinal, then it is called as <strong>Ordinal logistic regression</strong></li><li>If dependent variable is multi class then it is known as <strong>Multinomial Logistic regression</strong>.</li></ul><h3>3. Polynomial Regression</h3><p>A regression equation is a polynomial regression equation if the power of independent variable is more than 1. The equation below represents a polynomial equation:</p><pre>y=a+b*x^2</pre><p>In this regression technique, the best fit line is not a straight line. It is rather a curve that fits into the data points.</p><h3>4. Stepwise Regression</h3><p>This form of regression is used when we deal with multiple independent variables. In this technique, the selection of independent variables is done with the help of an automatic process, which involves <em>no</em> human intervention.</p><p>This feat is achieved by observing statistical values like R-square, t-stats and AIC metric to discern significant variables. Stepwise regression basically fits the regression model by adding/dropping co-variates one at a time based on a specified criterion. Some of the most commonly used Stepwise regression methods are listed below:</p><ul><li>Standard stepwise regression does two things. It adds and removes predictors as needed for each step.</li><li>Forward selection starts with most significant predictor in the model and adds variable for each step.</li><li>Backward elimination starts with all predictors in the model and removes the least significant variable for each step.</li></ul><p>The aim of this modeling technique is to maximize the prediction power with minimum number of predictor variables. It is one of the method to handle<a href="https://www.analyticsvidhya.com/blog/2015/07/dimension-reduction-methods/"> higher dimensionality</a> of data set.</p><h3>5. Ridge Regression</h3><p>Ridge Regression is a technique used when the data suffers from multicollinearity ( independent variables are highly correlated). In multicollinearity, even though the least squares estimates (OLS) are unbiased, their variances are large which deviates the observed value far from the true value. By adding a degree of bias to the regression estimates, ridge regression reduces the standard errors.</p><p>Above, we saw the equation for linear regression. Remember? It can be represented as:</p><p>y=a+ b*x</p><p>This equation also has an error term. The complete equation becomes:</p><pre>y=a+b*x+e (error term),  [error term is the value needed to correct for a prediction error between the observed and predicted value]</pre><pre>=&gt; y=a+y= a+ b1x1+ b2x2+....+e, for multiple independent variables.</pre><p>In a linear equation, prediction errors can be decomposed into two sub components. First is due to the <strong>biased</strong> and second is due to the <strong>variance</strong>. Prediction error can occur due to any one of these two or both components. Here, we’ll discuss about the error caused due to variance.</p><p>Ridge regression solves the multicollinearity problem through <a href="https://en.wikipedia.org/wiki/Shrinkage_estimator">shrinkage parameter</a> λ (lambda). Look at the equation below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/321/0*5gOFT2ju7xFWv9da" /></figure><p>In this equation, we have two components. First one is least square term and other one is lambda of the summation of β2 (beta- square) where β is the coefficient. This is added to least square term in order to shrink the parameter to have a very low variance.</p><h4>Important Points:</h4><ul><li>The assumptions of this regression is same as least squared regression except normality is not to be assumed</li><li>It shrinks the value of coefficients but doesn’t reaches zero, which suggests no feature selection feature</li><li>This is a regularization method and uses <a href="https://en.wikipedia.org/wiki/Regularization_(mathematics)">l2 regularization</a>.</li></ul><h3>6. Lasso Regression</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*UHo0M69AtTpUquXP" /></figure><p>Similar to Ridge Regression, Lasso (Least Absolute Shrinkage and Selection Operator) also penalizes the absolute size of the regression coefficients. In addition, it is capable of reducing the variability and improving the accuracy of linear regression models. Look at the equation below: Lasso regression differs from ridge regression in a way that it uses absolute values in the penalty function, instead of squares. This leads to penalizing (or equivalently constraining the sum of the absolute values of the estimates) values which causes some of the parameter estimates to turn out exactly zero. Larger the penalty applied, further the estimates get shrunk towards absolute zero. This results to variable selection out of given n variables.</p><h4>Important Points:</h4><ul><li>The assumptions of this regression is same as least squared regression except normality is not to be assumed</li><li>It shrinks coefficients to zero (exactly zero), which certainly helps in feature selection</li><li>This is a regularization method and uses <a href="https://en.wikipedia.org/wiki/Regularization_(mathematics)">l1 regularization</a></li><li>If group of predictors are highly correlated, lasso picks only one of them and shrinks the others to zero</li></ul><h3>7. ElasticNet Regression</h3><p>ElasticNet is hybrid of Lasso and Ridge Regression techniques. It is trained with L1 and L2 prior as regularizer. Elastic-net is useful when there are multiple features which are correlated. Lasso is likely to pick one of these at random, while elastic-net is likely to pick both.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/300/0*FaxMYReXynmymzWF" /></figure><p>A practical advantage of trading-off between Lasso and Ridge is that, it allows Elastic-Net to inherit some of Ridge’s stability under rotation.</p><h4>Important Points:</h4><ul><li>It encourages group effect in case of highly correlated variables</li><li>There are no limitations on the number of selected variables</li><li>It can suffer with double shrinkage</li></ul><p><a href="https://www.analyticsvidhya.com/blog/2015/08/comprehensive-guide-regression/">https://www.analyticsvidhya.com/blog/2015/08/comprehensive-guide-regression/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3417d077a3d3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[GHC’19 Experience(Orlando, Florida):]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-snippet">My friend went there last year, and she managed to get an intern at GoDaddy through that. Seeing it, I decided to invest in my carer a bit&#x2026;</p><p class="medium-feed-link"><a href="https://gunjanagicha.medium.com/ghc19-experience-orlando-florida-922bf80063e4?source=rss-aeb3f45c6182------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://gunjanagicha.medium.com/ghc19-experience-orlando-florida-922bf80063e4?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/922bf80063e4</guid>
            <category><![CDATA[tech]]></category>
            <category><![CDATA[grace-hopper-conference]]></category>
            <category><![CDATA[interview]]></category>
            <category><![CDATA[experience]]></category>
            <category><![CDATA[conference]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Sat, 05 Oct 2019 22:45:07 GMT</pubDate>
            <atom:updated>2020-03-01T19:59:31.570Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Face recognition]]></title>
            <link>https://gunjanagicha.medium.com/face-recognition-1c634c214fe9?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/1c634c214fe9</guid>
            <category><![CDATA[face-recognition]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Wed, 25 Sep 2019 03:48:23 GMT</pubDate>
            <atom:updated>2020-01-31T07:58:45.076Z</atom:updated>
            <content:encoded><![CDATA[<p>I am using face-recognition library which provides with a python interface and at the back uses dlib, openFace, OpenCV.</p><p><strong>Steps:</strong></p><ol><li>Face Detection: detects faces in the image given.</li><li>Face Encoding: convert that face image into numerical feature points that describe the face in the best possible way.</li><li>Comparison: Compare this feature vector to other faces vectors stored in the database. One that is closest to the input image vector and whose distance is less than the threshold is the predicted output.</li><li>Spoofing: Check if the face is real or a spoofed image.</li></ol><h4><strong>Step1: Face Detection</strong></h4><p>dlib’s HOG(histogram of oriented gradient) detector is used for this.</p><p>Hog detector checks for each pixel in the image the surrounding pixels to the current pixel. It then draws arrows in the direction in which the image is getting darker. In the end, the entire image is filled with arrows. These arrows are called gradients and they show the flow from light to dark across the image. We check this pattern to the earlier known face pattern of gradients and extract only the face part.</p><p>The original image is turned into a HOG representation that captures the major features of the image regardless of image brightness.<br>To find faces in this HOG image, all we have to do is find the part of our image that looks the most similar to a known HOG pattern that was extracted from a bunch of other training faces:</p><p>I have used the HOG face detection module, which is the default for face-recognition library’s get face location function. Cons: It doesn&#39;t detect very small faces. Not so accurate. Pros: Fastest of all implementations. For HoG detector face size&gt; 80X80</p><p>Another option is to use DNN which uses Resnet to detect the face. Pros: Its the most accurate method. Can even detect very small faces, and side face angle. Thus DNN can be used if latency is not an issue and accuracy is of utmost importance.</p><h4><strong>Step2: Encoding Face</strong></h4><p>Now instead of comparing the entire face to al the faces in the database, we only compare some extracted features from the unknown face to all the known faces. For example, we might measure the size of each ear, the spacing between the eyes, the length of the nose, etc.</p><p><strong>The most reliable way to measure a face(Triplet loss)</strong></p><p>Ok, so which measurements should we collect from each face to build our known face database? Ear size? Nose length? Eye color? Something else?</p><p>It turns out that the measurements that seem obvious to us humans (like eye color) don’t really make sense to a computer looking at individual pixels in an image. Researchers have discovered that the most accurate approach is to let the computer figure out the measurements to collect itself. Deep learning does a better job than humans at figuring out which parts of a face are important to measure. We will train a deep neural network to generate 128 measurements from each face.</p><p>The training process works by looking at 3 face images at a time:</p><ol><li>Load a training face image of a known person</li><li>Load another picture of the same known person</li><li>Load a picture of a totally different person</li></ol><p>Then the algorithm looks at the measurements it is currently generating for each of those three images. It then tweaks the neural network slightly so that it makes sure the measurements it generates for #1 and #2 are slightly closer while making sure the measurements for #2 and #3 are slightly further apart. After repeating this step millions of times for millions of images of thousands of different people, the neural network learns to reliably generate 128 measurements for each person. Any ten different pictures of the same person should give roughly the same measurements.</p><p>I am using OpenFace’s library ResNet based pre-trained neural network to get 128 measurements from each face.</p><p><strong>Step3: Comparison</strong></p><p>We check our face-encoding to the encodings stored in the database. I have used Euclidean distance to check which is the closest match.</p><p><strong>Step4: Checking spoof</strong></p><p>The image can be of a person or an image of the person’s image, i.e it can be a spoof. So to increase security I added a spoofing layer.</p><p>I used in-house employees data to recognize if the image is real or fake. A 4 layer CNN network was trained using Adam optimizer minimizing cross-entropy loss. I got a testing accuracy of 80% with just 50 training examples.</p><p>I have used facial recognition using one-shot learning by a deep neural network.</p><h4>What is one shot learning?</h4><p>In one shot learning, only one image per person is stored in the database, which is passed through the neural network to generate an embedding vector. This embedding vector is compared with the vector generated for the person who has to be recognized. If there exist similarities between the two vectors then the system recognizes that person, else that person is not there in the database. This can be understood by the below picture.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a5TeCJTxa_xkM8WQFaT63w.png" /></figure><p>Code: <a href="https://github.com/gagicha/Face-recognition">https://github.com/gagicha/Face-recognition</a></p><p>References:</p><ul><li><a href="https://face-recognition.readthedocs.io/en/latest/readme.html">Face Recognition - Face Recognition 1.4.0 documentation</a></li><li><a href="https://medium.com/@ageitgey/machine-learning-is-fun-part-4-modern-face-recognition-with-deep-learning-c3cffc121d78">Machine Learning is Fun! Part 4: Modern Face Recognition with Deep Learning</a></li><li><a href="https://cmusatyalab.github.io/openface/">OpenFace</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1c634c214fe9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[LINEAR REGRESSION]]></title>
            <link>https://gunjanagicha.medium.com/linear-regression-d7fd41d35c84?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/d7fd41d35c84</guid>
            <category><![CDATA[regression]]></category>
            <category><![CDATA[predictions]]></category>
            <category><![CDATA[linear-regression]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Tue, 24 Sep 2019 21:20:44 GMT</pubDate>
            <atom:updated>2019-09-24T21:20:44.298Z</atom:updated>
            <content:encoded><![CDATA[<p>We try to find the best fit line/plane that represents the relationships between i/p and o/p. This best plane is built by learning the parameters of a function that maps i/p to o/p.</p><p>E(y)= B0 + B1*X</p><p>Hence, if we find B0 and B1 parameters, we can predict our output. So Machine Learning basically tries to guess these parameters which minimize the cost function.</p><p>cost function: Mean Square Error</p><p>Why do we square them? It penalizes the values which were further away from the actual value.</p><p><strong>Aim: find a linear model that minimizes the MSE.</strong></p><p><strong>Assumptions: There is a linear relationship between the i/p variables and o/p.</strong></p><p>Correlation coefficient: How strongly the two things are correlated.</p><p>So for a linear model, a high correlation between dependent and independent variable means they are strongly linearly correlated.</p><p>How to make sense of the equation:</p><p>for every 1 step increase in x, we expect y to increase by B1</p><p><strong>Multicollinearity:</strong> When there are many independent variables, these variables might be related to each other, this is called multicollinearity.</p><p><strong>Null Hypothesis(H0)</strong>: the Currently accepted value of a parameter</p><p>Null Hypothesis for regression: There is no relationship between x and y.</p><p><strong>Regression statistics:</strong></p><ol><li><strong>r</strong>: coeffiecient, describes the relationship between each independent variable and dependent variable.</li></ol><p>correlation coefficient given by Pearson coefficient:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/530/1*NeyQI2mQffVCuHFJ_Zy6Ng.png" /></figure><p>or</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/286/1*358kWeDc3A-Xu2DZ3JYaoA.png" /></figure><p>2.<strong> r²</strong>: <em>coefficient of determination</em></p><p>represents the percent of the data that is the closest to the line of best fit. For example, if <em>r</em> = 0.922, then <em>r²</em>= 0.850, which means that 85% of the total variation in <em>y</em> can be explained by the linear relationship between <em>x </em>and <em>y</em> (as described by the regression equation). The other 15% of the total variation in <em>y</em> remains unexplained.</p><p>3. p- value: We use p-value to see which terms are significant and should be kept in the model. if it&#39;s less than .05, then you can reject the null hypothesis.</p><p>4. standard error: how far away the points are from the predicted line.</p><p>5. t-stat: coefficient/standard error</p><p>6. Confidence: How confident are we in our decision. If u are more than 95% sure, thats considered to be good.</p><p>Significance: 1- confidence. So significance&lt;5% is good.</p><p><strong>Cross-validation:</strong> helps to evaluate machine learning models by choosing the best model with minimum error. We keep some part of the training data as testing data on which error of each model is calculated. The best model is then passed through the actual testing data to determine the testing accuracy. This also makes sure that the model does not overfit.</p><p>Types:</p><p>a) Hold out: prone to sample bias</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/868/1*Cwk8-Jr-iaK7ryr39C2mgA.png" /></figure><p>b) k-fold: divided training data into k sets, train model on k-1 sets and 1 set as the validation set. Take the average of all the errors. Selection bias will not be present.</p><p>c) leave one out CV: a special case of k-fold validation where k=no od data points. 1 data point is taken as a validation sample each of the n cycles. then take an average of this error. con is that it will take lot of time.</p><p>d) bootstrap: randomly draw N samples from the training sample with replacement, and train the model on this set, while testing it on the left-out data points. The average over each test set error.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/891/1*i6DoUiqqZHyKXO2XrvPBQg.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d7fd41d35c84" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[1 month into college]]></title>
            <link>https://gunjanagicha.medium.com/1-month-into-college-d32298b862de?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/d32298b862de</guid>
            <category><![CDATA[masters-degree]]></category>
            <category><![CDATA[dallas]]></category>
            <category><![CDATA[life]]></category>
            <category><![CDATA[college]]></category>
            <category><![CDATA[computer-science]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Wed, 11 Sep 2019 20:12:55 GMT</pubDate>
            <atom:updated>2019-09-11T20:58:12.191Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Life in college is super busy!!</blockquote><p>That’s what I heard from my friends who went to masters last year. But, it’s a pleasure for me to be able to witness it. There’s so much to do and I have to prioritize my to-do list each morning.</p><p>Its been not even a month since my college started, but I have done and achieved so much in the last few days then I earlier did in 3 months.</p><p>A few important things that I did during the past month are:</p><ol><li><strong>Making resume:</strong> I attended a resume and interview workshop that are compulsory to attend my college. I asked a few of my friends and seniors to proofread it for me. So after 2–3 days my resume was good to go.</li><li><strong>Upload resume on GHC database:</strong> I uploaded my resume to the GHC database, and it got converted into a plane word format on their site. I had to edit it again to look somewhat decent.</li><li><strong>Cover Letter:</strong> Many of the applications have an option to upload cover letter too. Even though it’s not a required option, I feel one must upload it. It took me 2 hours to prepare a cover letter, got few friends to proofread it. I keep updating it before uploading it to a new company.</li><li><strong>Updating LinkedIn:</strong> It took me a day to update my LinkedIn profile, I uploaded my resume, added content in the About section, added experience, skills, awards section. Added a formal picture as DP.</li><li><strong>Make a profile on Indeed and Handshake:</strong> Handshake is used in our college UTD for on-campus jobs too. So I made profiles on both the sites.</li><li><strong>Connect with recruiters:</strong> I searched on LinkedIn technical recruiters/ university recruiters for the companies where I want to work and sent them a connect request with a note saying who I am, what I am doing etc, and that I want to connect with them regarding any opportunities.</li><li><strong>Applying to companies: </strong>I am particularly looking for data science internships positions, so I created a job alert for it. Each morning I will get an email for these positions, and I apply in it. It’s important to create an alert, as its best to apply as soon as the position opens and be among the top few candidates. Each application directs you to their website, where we have to create our profile and enter many details, upload resume, and cover letter, etc. Each application on average takes about 8–10 mins. So, Every night before sleeping I used that time to apply.</li><li><strong>Keep the excel sheet of the companies applied: </strong>I maintained an excel sheet of the companies I applied to, with the status of the application, the platform I applied from, date applied on, etc. This makes tracking companies easier.</li><li><strong>Leetcode Practice: </strong>I had started doing Leetcode problems last year because my friends suggested me to start early. But, after coming to college when I started it again, I fell blank. I was struggling to write even the easiest question of all. So, I started practicing it every day. And I made sure to give 1 hour every morning for it, as at the end of the day after assignments and classes, there will be no extra time for this. So I gave a priority to it. Everyday 2–3 easy problems is what I am doing. Eventually, I will shift to medium, once my speed and accuracy increases in easy problems.</li><li><strong>Courses: </strong>I have 3 courses this sem, meaning 9 hrs every week of classes. Doing assignments, studying for it takes much more time. Like I have data structures course, and prof wants the cod only in Java, which I am not good at. Hence every week it’s taking me around 5 hrs to study its concept, before even starting the assignments.</li><li><strong>Coding Challenges: </strong>I am going to Grace Hopper this year, and have uploaded my resume on their database. I have got 2 coding challenges from it till now. One from point72, and another from Salesforce. So each challenge for 1 hr. I am hoping to get more before and after GHC.</li><li><strong>Machine Learning: </strong>I am looking for data science internships, if I get a call I have to be prepared for the interview. I don’t have the ML course this sem, and hence I have to revise it on my own. So, I am referring a few ML books for it.</li><li><strong>Lyft Data Science Challenge: </strong>I participated in their intern data science challenge, and it’s tough. I did some analysis but have not figured it out fully. Its really time consuming, it says it will take 8 hrs, but that’s already consumed and I am not even half done. We have to code + give a writeup to them.</li><li><strong>FB event: </strong>FB came on-campus at UTD this year. I attended the event and gave my resume to the recruiter, I told her about my experience and that I am going for GHC. She kept it at the top of all the resumes till the end, I might get a call from them, or I will go meet them during GHC.</li><li><strong>Applied to college GHC scholarship</strong>: UTD gives out a few scholarships for GHC, so I applied for it. It was a 1-page form that I had to go and submit at an office.</li><li><strong>Sports:</strong> College is tiring, and I want to have fun too. Hence, once or twice a week, I take out time to play BB, swimming, or squash. Its a good refresher for me.</li><li><strong>Cooking:</strong> Its the most difficult thing to do in college. At home when we were in school, our moms made sure to cook for us. But now, it’s our responsibility to cook every meal and manage all the above things in our life.</li><li><strong>Home Groups:</strong> There are many home groups at Texas, and I am linked with one of them. Every Friday it happens at one of the families houses, they cook us dinner, we play some games and meet new people. Once a week, it feels good to take a break from cooking and just go and have fun.</li><li><strong>Grader position: </strong>I applied to the grader position in the computer science department and managed to get it. So I have to work for 15 hrs per week on that.</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d32298b862de" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Feeling depressed vs Feeling happy]]></title>
            <link>https://gunjanagicha.medium.com/feeling-depressed-vs-feeling-happy-2101714690a7?source=rss-aeb3f45c6182------2</link>
            <guid isPermaLink="false">https://medium.com/p/2101714690a7</guid>
            <category><![CDATA[college]]></category>
            <dc:creator><![CDATA[Gunjan Agicha]]></dc:creator>
            <pubDate>Sun, 11 Aug 2019 13:40:07 GMT</pubDate>
            <atom:updated>2019-08-11T13:40:07.068Z</atom:updated>
            <content:encoded><![CDATA[<p>I am writing this blog after I have been through both phases. So I know how it feels in both the stages.</p><p>First let me talk about my experience with depression, and how I dealt with it. In December 2018, I applied to 8 colleges for my master&#39;s degree in both the US and Canada. I waited 3 long months for the results to come. And when it came in April, the first result was a reject. I cried a little. But then the second result came, and… it was also a reject. Results kept coming in one by one across the month of April, but all were rejects. I lost all the hopes of getting an admit. It was a tough period to go through. One loses all the confidence in themselves, I felt like it was my fault. In general, getting an admit is getting tougher each year, as students applying to these universities keep increasing. And also the admission process is very random, so one cannot judge themselves based on it. One of my friends applied to just one college and he got into that, despite having provided with only one letter of recommendation (when asked LORs were 3). I felt depressed with each passing day. By this time, I had planned to put down my papers in April, and here I was with my plans failed. I was really unhappy with my life at that time. I was sleeping more than my usual 8 hrs. I didn’t feel like getting up and facing another day. Yeah those lows are usual with people who are unhappy with their life. I stopped doing any exercise during this waiting period. And lastly, I was talking very less to my parents on calls, as they would ask if I received anything, and I would feel bad again. This was my life for an entire month. On the work side, there was not much going on in the office, so many days I would just connect from home and watch some series. So basically, I was living a life with no purpose and felt bad all the time. A friend of mine pushed me to start executing a Plan B, which was basically to switch jobs in India in my field of interest i.e. Machine Learning. He told me that eventually, I would be searching for a new job, so start early and get out of my dismal state. So, I started studying for companies, but I found that I was not giving it my 100%, I was not even giving it my 50%. I still had 2 colleges to hear from, and although I was disheartened by all the rejections, there was still hope. Hence, my efforts to change a job and study for it were half-hearted. During this time, I didn’t shut myself completely, I used to hang out with friends on all weekends. We all need to be with people who we feel our best with. That helped me to not feel alone.</p><p>Now the part where happiness poured in my life. On 30th April, I got an email from one of the remaining colleges and it had an attachment (rest all mails had been direct text saying we regret to inform you). I opened it, and there it was, an acceptance email from the University of Texas, Dallas. I called up my family and friends and shared the news. They have heard me cry so many times for all the rejects. Finally, here I was with an admit and I would be going for Masters( provided Visa gets done). After admits there were lots of things that needed to be completed, loan process, getting my i20 and applying for Visa. This kept me occupied and happy. I was talking to my parents often now and sharing all the details of my day. I slept well, ate well and started exercising again. I just felt happy! All the tension was gone. Now I used my time to prepare myself for college, learning to drive, working out every evening, having dinner and relax watching Netflix. And I just feel content and happy with it.</p><p>So bad days will eventually pass, dealing with it is tough but there is no other option. Keep living it day by day, surround yourself with people you feel most comfortable with. Try to push yourself to do something, do anything. A small achievement during this low time, makes one feel happy.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2101714690a7" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>