Types and Tailcalls

Ultralearning Data Science - Week 9

published on December 23rd, 2019

How the Ninth Week Went

The ninth week went different from how I thought it would. I was off from work, so in theory I should have had a lot of time to work on the project but in practice, life interfered and I had a lot of other things to take care of which let me spent almost no time on the ultralearning project. I'm the first person frustrated by this, but I also want to keep this free time project from frustrating me too much.

What I did do during this last week was to complete the first deeplearning.ai course, progress reviewing notes of the Andre Ng's first ML course and read the chapter on backpropagation of the neural networks and deep learning book (NNDL) and the backprop post on Colah's blog. To be honest, I found the presentation of the backprop algorhim to be not that great in the deeplearning.ai course and also needlessly complicated in the NNDL book. I liked Colah's the best, but it is a bit generic and doesn't focus really on backpropagation in neural networks. I'm planning on writing my own blogpost to rectify this situation 🤪 (yes I'm fully aware that a million blog posts have been written on this topic and most are better than mine will be, I'm doing this for my own education)!

Over the next weeks, I'm still off from work but have a lot of family committments. I've decided to not explicit goals for what I'm working on (which I'm not meeting much anyway) and instead work on what I find most appropriate at the time. Let's see how this works.

Reviewing Goals for Week 9

Looking back at my goals for week 9, here are the results. Yeah, I'm not happy with them, but they are what they are:

  1. [incomplete] Complete review of Andrew Ng's ML course, create flashcards
  2. [incomplete] Complete the first deeplearning.ai course with notes and flashcards
  3. [failed] Complete lectures 2-4 of the fast.ai course with notes and lectures
  4. [failed] Complete NLP experiments on my own text classification project
  5. [failed] Do some practical experiments on the SVG project.
  6. [failed] Implement Neural Network from scratch and train it on MNIST data.

Things I Might Work On

As stated, I'm giving up on goals for the next week and try indeed to make as much progress as I can on whatever I feel most like working on at the time. Here is a list of things which I currently think are the most interesting for me:

comments powered by Disqus