Hacker News Reader

Hacker News

Comment by melling | Hacker News Reader

Full thread

melling

I’m 26 weeks into doing machine learning on Kaggle every day. Slow and steady practice.

https://www.kaggle.com/mmellinger66

I’m also part of a small Discord study group that is learning ML theory. We’ve gone through this book completely:

https://mml-book.github.io/

Just finished the foundations part of Murphy

https://probml.github.io/pml-book/book1.html

Two of us have started to independently study the MIT Probability course, while working through the problems together

https://ocw.mit.edu/courses/6-041sc-probabilistic-systems-an...

Study groups often start out with 6-10 people but if you can find only 1 good person who you work well with, it’s extremely motivating.

View on HN

How'd you get the study group started? Some peers you already knew? I've previously considered trying to gather some people to regularly read/practice other books/topics but wasn't sure where to begin gathering others.

It’s extremely difficult. Through Meetups, Discord, …

A group of us made it through Hands on ML last year.

https://github.com/ageron/handson-ml3

But that group didn’t want to do heavy theory but someone from another Meetup did so I invited them to our group.

Then someone who didn’t want to do the theory changed their mind.

Try meeting people online with a well-defined common goal.

For example, not simply “I want to learn machine learning” but I want to study a particular book or MOOC.

Getting people to agree can be difficult. No one else in my group wants to do Kaggle, for instance. Lots of cool project ideas that don’t go anywhere.

Discord seems like it should be the perfect place to meet people but most servers aren’t very active.

Here’s an OpenCourseWare server I found a few days ago that’s disappointingly not very active:

https://discord.gg/CSKeQegY

Ideally math, physics, writing, history, or whatever could be served by dedicated servers.

Thanks for the tips. It seems like a good idea then to consider a specific book and go about gathering others from there, so that everyone has the same expectations of content.

yeah, i am about to wrap up the MIT probability course. Very very good course, i will be happy if i retain 50% of what i learned. Markov Chain is my favorite so far.

Harvard Stat 110 might be an interesting follow up in a few months.

https://projects.iq.harvard.edu/stat110/home

Videos on YouTube and his book is available on Gdocs

Reply on news.ycombinator.com