Naveen Mathew Nathan S.

Nov 21, 2022

Between a job application and an interview - data science / machine learning tech jobs

This is a follow up to my previous article on applying for tech jobs (data science / machine learning job application processes are similar). This artcile will outline the different types of interview processes. Note: This article is slightly opinionated because of my strong inclination towards “process justifying the results”…

Job Hunting

4 min read

Job Hunting

4 min read

Nov 18, 2022

Applying for technology jobs

This article aims to document important aspects of my job search in the USA. This article will focus on online job applications only (I’m discounting direct approaches from recruiters and other modes to match the current job market). Note: I was extremely lucky to sidestep the hiring slumps. My hope…

Job Hunting

3 min read

Job Hunting

3 min read

Aug 15, 2022

Manufacturing polynomials using a sigmoid neural network — practicum

In the previous article I discussed a possible way to create polynomial bases (plural of basis in linear algebra) to approximate a polynomial function using a sigmoid neural network. This article is a practical test of the idea — does the bias constraint work as predicted, and how does it…

Deep Learning

4 min read

Manufacturing polynomials using a sigmoid neural network — practicum

Deep Learning

4 min read

Dec 29, 2021

Manufacturing polynomials using a sigmoid neural network

The objective of this article is to understand how (deep) neural networks with sigmoid activation can manufacture polynomial-like functions. The objective of this article is to only “engineer” possible bases (plural of basis in linear algebra) for higher order polynomials in the hidden layers to approximate any polynomial function using…

Deep Learning

6 min read

Manufacturing polynomials using a sigmoid neural network

Deep Learning

6 min read

Dec 28, 2021

Understanding the Expressive Power of ReLU Networks

Introduction This article was inspired by Wilson Wang’s article - Understanding the Expressive Power of Neural Networks. Wilson’s original article is applicable to a wide range of activation functions. The focus of this article is to create a spline equivalent of a single hidden layer neural network that uses ReLU activation…

Spline Regression

7 min read

Understanding the Expressive Power of Neural Networks — ReLU Activation

Spline Regression

7 min read

Oct 31, 2021

Does python prioritize True in ‘or’ and False in ‘and’ condition checks?

Background I found this meme on LinkedIn recently. After reading this post I hope people feel the same about python. Experimenting with ‘and’ Let’s assume we have a simple function in python to check if all the numbers in a list are greater than a threshold. Logically the function should return False if any…

Python

4 min read

Python

4 min read

Published in Towards Data Science

·Oct 5, 2019

Ridge regularization on linear regression and deep learning

Introduction As discussed in my previous articles on regularization and lasso penalty, regularization can be used to counter overfitting in overparameterized models. Regularization serves as a computationally efficient alternative to best subset selection, but has its disadvantages (Eg: efficiency of the estimator is low). …

Machine Learning

4 min read

Ridge regularization on linear regression and deep learning

Machine Learning

4 min read

Published in Towards Data Science

·Sep 30, 2019

Lasso regularization on linear regression and deep learning

A mathematical analysis of the effects of lasso penalty and its effects on linear regression, including possible extensions to deep learning — Introduction Curve fitting — under and over fitting As discussed in my previous article, issues with ‘curve fitting’ occur when the problem is ill-posed. Underfitting is usually not a big problem because we have the option to expand the feature set by acquiring/engineering new features. However, overfitting is not easy to handle.

Machine Learning

5 min read

Lasso regularization on linear regression and other models

Machine Learning

5 min read

Published in Towards Data Science

·Sep 21, 2019

Lasso, ridge and dropout regularization — their effects on collinearity

Introduction This is a follow-up to my previous amateurish and cryptic article — A short note on regularization. The article delivered what it promised, but it is insufficient to answer the questions — what does regularization do and why does it work when the corresponding model without regularization doesn’t? …

Machine Learning

6 min read

Different forms of regularization and their effects

Machine Learning

6 min read

Sep 21, 2019

Search for exoplanets — humans vs stars

Kepler Kepler telescope was launched in the year 2009 into an Earth-trailing heliocentric orbit. The planned lifetime of the mission was 3.5 years, but the mission continued for almost 10 years. The objective was to scan a portion of the sky to identify Earth-like habitable exoplanets in a region of the…

Astronomy

5 min read

Astronomy

5 min read