How was my Winter with code?

This winter was quite busy and filled with projects! I will go through most of the new things that I explored.

Sublime -> Visual Studio Code

I switched my laptop from a 5 year old Dell Inspiron to Mac Air and now have 16GB RAM, enough to get a heavier text editor. VS Code, although heavy, is an amazing text editor. In-built git support, wonderful plugins that are easy to install and great debugging support made me smile several times. Meanwhile I also learned Vim, but the pain in setting it up caused me to stay with VS Code and use my Vim knowledge when I ssh to a cloud instance.

More RAM

16GB ram is so wonderful for someone who used laptop with 4GB ram for several years for Web dev and Data analysis. Now I am using cloud instances less often, as most competition data can fit easily in the ram. So much convenience! You have more? *shush* 😮

Freelancing

I was low on ideas to implement something so I tried freelancing on Upwork for Machine Learning projects. The projects were short but turned out to be very beneficial. I got to implement several Deep Learning models for time series data. The data was mostly related to medical science, so I also got to research about brain disorders and their symptoms. I had to identify these disorders using various kind of brain signals.

I might write an article in future covering signal processing and using deep learning on that type of data.

Golang

I learned Golang! It is such a beautiful, light and interesting language. Its lack of generics and operator overloading still bothers me, but the concurrency control in Go filled up all the cons.

Learning Go is easy, you can get started with the go tour,

go get golang.org/x/tour/gotour

I still have to master the concurrency patterns, implementing web applications with Go.

Chingu now has a new Clustering Algorithm!

Recently I revisited the Chingu Sort project, earlier it was meant for just finding similar members on basis of timezones, and I didn’t have much knowledge about using the text as a feature too.

Now I have built a nice Scikit Pipeline based model which I might be covering in a future article soon! We used it for Voyage 4 and it was able to identify similarities in several dimensions. To be honest, too many dimensions!

I’ll be nerfing it down to give it more focus on what features are necessary and what aren’t.

Google Summer of Code 2018!

GSoC is back and with it the opportunity to contribute to lots of open source projects. Open source code is always a great way to build up skills. Last year I got rejected but even in the 2 months of contributions I learned a lot about C++ and Evolutionary strategies. Which I wrote about in this article.

So I am going to try to get in GSoC this year again. This year I arrived early so I was able to grab some Pull Requests and discuss with mentors easily. This is important because as time progresses, they become more and more busy.

If you are new to the GSoC craze, just follow these steps:

1. Find an organisation and project idea (provided by them) that you would love to contribute to.
2. Talk with the mentors and discuss what might be a great proposal for them.
3. Contribute some PRs before the real thing starts to show dedication.
4. Rest is on luck. All mentors are kind enough that if they have enough seats for GSoC, they’ll choose all applicants! But due to the competitive nature of this event, they have to tie their hands.

Rejection can hurt, but the open source community is always open to give hugs and teach you. 😄

Hands on Machine Learning with Scikit-Learn and Tensorflow

This book is amazing and gives great mathematical intuition about the working of various algorithms! I got it for free in the O’Reilly publication giveaway, and read it over a few weeks. This book gave me a lot of information about creating Pipelines, distributed Tensorflow and approaching a feature engineering task.

The explanation of autodiff algorithm in Tensorflow greatly helped me understand using the graph api more efficiently and organised manner. This was the best and informative section. I wish it covered GANs too as they can be quite confusing to understand but they are a big breakthrough in DNN architectures.

I highly recommend this book when you are comfortable with ML and want to go deeper into understanding Tensorflow.

Clean Code

This book by Uncle Bob is amazing! It helped me understand that code doesn’t need comments to be expressed. Now whenever I write some code, I try to make the names and classes as expressive as possible. Over the course of the book, you are shown code, the dirtiness of it and how it can be more expressive.

Highly recommended whatever coding skill you have, this book will just help you understand clean code.

Some images from this journey!

I tried learning Blender :P

Received my Hacktoberfest shirt!

And the stickers! :D

The mlpack sticker is for my GSoC 2017 contributions in gym-tcp tool, Thanks mlpack! :D

Analysing Chingu Applications with Word clouds!

Word clouds are a great way to have an overview of text data, I’ll show creating and using them properly in the article for presenting Chingu’s new clustering model.

That’s all folks! Thanks for reading :) I hope this journey keeps reaching new milestones and so does yours!