Types and Tailcalls

Ultralearning Data Science - Week 6

published on December 1st, 2019

How the Sixth Week Went

The sixth week of my learning project was once more characterized by quite a few social commitments which limited the amount of time I was able to put into it. As it stands I spent a little over 9 hours on the project, a long cry from my planned 15 hours.

This week, I've continued to feel that I'm more comfortable with basic data science questions and can navigate strucutred data with pandas with greater ease. I definitely still have a lot to learn on the pandas front, but at least I can now get basic things done and figure out the rest with a little more time. It feels like I've reached or am close to basic competency in this field.

I've also had some interesting ideas on how to improve note-taking and about a better approach to learn pandas - to be honest I think 6 weeks is a bit long to feel comfortable navigating the library and doing basic data analysis.

What Went Well

Exploratory data analysis: I'm happy to say that I've completed my first public exploratory data analysis for the 2019 data science bowl. I'm not exactly proud of the result but it is something useful and I've completed it. I am kind of proud of my time management, because my natural instinct would have been to keep going and dig deeper and deeper, but I was aware that I had no chance of meeting my other goals if I did that and so I shipped something that was minimally useful.

Started reviewing lecture notes from Andrew Ng's ML course: I haven't completed this (that was way too ambitious), but I've continued reviewing the lecture notes from Andrew Ng's first ML course and created Anki cards for it. I do think this course is very valuable and I want to continue doing this.

What Went Badly

Didn't complete CUDA setup and first neural net: This should have been short, but I stumpled over some library incompatibilities that I couldn't resolve with the time I had. I will keep trying this next week.

Didn't review all Fast.ai lectures: This was completely too much, I put this back up on my calendar for next week.

Didn't complete reviewing the lectures from Andrew Ng's ML course: This was also too much, though I could have done a bit more than what I did.

Lack of discipline: Sometimes I struggle to find blocks of time and work uninterupted on the project. I want to get more out of the time that I put in. I'm generally quite good (and happy with) how I stay on this project without major distracitons, though.

Reviewing Goals for Week 6

Looking back at my goals for week 5, here are the results:

  1. [done] Complete the EDA for the data science bowl 2019 Kaggle challenge.
  2. [done] Watch lectures 9 and 10 of fast.ai.
  3. [failed] Build my own neural network based on the content of the fast.ai lectures.
  4. [failed] Review fast.ai lectures so far and write Anki cards.
  5. [started] Review lecture notes from the ML coursera course by Andrew Ng and write Anki cards.

My Goals for Week 7

This is my last week of work at my current job and I have two days where I will not be working. The last days at my job will be somewhat busy, so overall for this week I think I should be able to put in a little more time (goal 20 hours) than during the last weeks. At the same time, I'm about to complete the lectures from the Introduction to Machine Learning fast.ai course. I want to combine these two events to review and restructure my notes before starting of onto a new angle next week.

My goals are as follows:

  1. Get the neural net notebook to work.
  2. Watch lectures 11 and 12 of fast.ai.
  3. Review fast.ai lectures, consolidate and publish meeting notes and write Anki cards.
  4. Keep reviewing lecture notes from the ML coursera course by Andrew Ng and write Anki cards (no completion required).

comments powered by Disqus