This is a guide to make deep learning less messy and hopefully give you a way to use less tissues next time you code.

Sad robot image
Sad robot image

If you can answer yes to most of them. Read. Or cry. Your choice of course.

  • Do you work with deep learning models?
  • Why are there 15 hyper parameters???!!!!
  • Frequently run into bugs?
  • Wish you didn't have to open google every 5 minutes while coding?
  • I just want a working model dammit!!!
  • Please. Why am I getting 3% accuracy. It's been 2 days. Please.
  • I should have done an MBA. Why am I here. What is…

dog sleeping ; Photo by Priscilla Du Preez on Unsplash

Pull to refresh. Swipe to refresh. Swipe up. Scroll. Scroll. Pull. Scroll.
Sounds like a typical day? Me too. But why does it all happen? Why do we keep doing the same things, looking for the same things to happen. Even more so these days when almost all of us are stuck at home. I’m no psychologist by any means but lately, I have been thinking a lot about this and so I was prompted to share my story and ways I have found that helps.

Over the past year or so, a typical routine for me would involve the…

Frustration -> Happiness from

Ever made a library, added documentation and then thought “Ugh I need to make this into a website so my users can see whats there. Why is it so much work?”. Here is the solution to your problems. (No fixes for demotivation sorry)

Note: This only works for Python code. It will work perfectly on Linux/MacOS/WSL with no issues. Windows.. you might need to change the commands to suit your shell.

What you will have by the end of this short tutorial

  • An interactive website with all your modules, sub modules, classes documented nicely. It will even show code!
  • No…


Over the past few centuries, civilization has progressed rapidly but at the cost of our environment. Illegal deforestation as well as hunting is at an all-time high due to the current situation and human greed. Can we make a humble effort to save our forests using deep learning? How about trying to identify if any sounds were corresponding to that of logging or shooting?

Let us see how we can implement this using PyTorch in a simple to understand manner.

> This article has been greatly inspired by Blog 1 and Blog 2


Before we get into any of the…

Do you use social media? Ever sent someone a picture and when you look at it later you find that the image is so bad in quality? What if you could reverse that?

Link to image

What if

The main objective -> Take an image and upsample it by 3x (That means if an image is initially 100x100 by the end we have a 300x300 image) without losing any information. Now if we decided that we want the image to be 100x100 we could resize it but it would be of a much higher resolution than before.

We will look at how to pose this…

Distributing ; [Link]

Want the smarts machine learning offers? That is awesome, but how secure is your data there really? And what if you could have the benefits without there being a chance of your data being leaked? We will have a look at Federated Learning and understand how it works. We will also look at a few papers and what they have to say on the matter and take the library Syft for a whirl.

Why is this a problem?

Think about this from the perspective of an app that connects you with your doctor. It is a great app and even can forecast future appointments and…

Why we ignore the magic of some layers in a neural network, and how to understand them better + 9 takeaways from an extremely interesting paper.

A batch of flowers (

Batch Norm is one of the most widely used layers in a neural network. Ever since it came out, it became possible to train neural networks that were faster, more accurate, and more resistant to change. Sounds almost magic doesn’t it? You would think, for something so magical, the implementation must be crazy hard. (You would be wrong)

What happened is that due to the black-box nature of a neural network, we started taking…

How smart are neural networks? And can we break them and fool them into doing dumb things?

Our friend Termy


Today we have a special guest. His name is Termy, he is a friendly robot from the year 2025. What is cool about him is that he has an AI and a really nice personality. He loves people and animals. (Or at least is told to). He is programmed to never hurt them. The only thing he hates is guns. He will try to break them as soon as he finds them.

On the way sadly, his system was hacked. Termy, being from…

Pruning is something I have been interested in for a long time but somehow I could never get around to implementing it. It interested me for a lot of reasons. Mainly that of being able to reduce the size, cost, and computational requirements of my models, all while maintaining the accuracy (sort of at least).

TL;DR Generally this comes about by removing parameters in some form or fashion.

Rather than taking a mask, we can prune certain parts of the network by setting them to 0 or by dropping them if required. (aka weights and biases)

In most cases, the…

Dear Researchers,

The lockdown is over and you are standing in a line to the movies. You can feel the excitement, feel your pulse racing. Yes! This is going to be amazing. Due to high demand, seats are first come first serve. A member of the staff comes and screams that everyone can go in now. Suddenly there is absolute chaos.

7 doors, a few hundred people. Everyone starts running. In their midst, you stand paralyzed. Which door do I pick? What if… I don’t get in? By the time you start running.. it is too late. You have missed your chance…

Subhaditya Mukherjee

I am a dreamer and coder. Using my computer to get my thoughts to reality and trying to make the world better, one smile at a time :)

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store