Yugandhar_Master's blog

Suggestions to start Machine Learning

By Yugandhar_Master, history, 5 months ago, In English

Hello community, I want to start to learn machine learning, so please give me suggestions and good resources and the mistakes which I need to avoid in learning process. Thank you

  • Vote: I like it
  • 0
  • Vote: I do not like it

»
5 months ago, # |
  Vote: I like it 0 Vote: I do not like it

I don't think this is the best place to ask this. A lot of people here hate anything AI/ML

→ Reply
»
5 months ago, # |
← Rev. 2   Vote: I like it +10 Vote: I do not like it

pytorch.org has very good tutorials, and i'd recommend looking at those and coming up with very similar but slightly modified tasks and trying to do them, start simple and build from there. I am still very noob, but reading and implementing several of those along with reading corresponding and similar papers helped me have much better insight i think. also andrej's blogs are nice overviews.

If you know nothing about neural nets and how backprop works, i'd also watch the 3b1b videos, make a small example with paper by hand, and glance through video and repo/maybe implement andrej's micrograd. just look up and learn the linear algebra you need as you go.

→ Reply
  • »
    »
    5 months ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    i know kaggle exists, but i hope someone will make ML website with problems more in style of competitive programming — problems that require clever twists on known ml methods and you pass if you get above some accuracy. maybe i will make one day but probably not...

    → Reply
»
5 months ago, # |
  Vote: I like it 0 Vote: I do not like it

There is a book "Understanding Deep Learning" by S. Prince. I think that is quite nice for conceptual understanding. I also think its important to get your hands dirty by coding up stuff. So learn pytorch and implement random stuff. Like train a models to do sentiment analysis on some random data from your personal life, image classification, play simple game, do a kaggle competitions etc.

→ Reply