Give me tips to learn ML

Beginners
1

I have a few questions, so I can have a basic idea about how to learn, thanks!

2 Likes
2

Anyway, let’s actually ask the question.

3

How can I begin learning?

2 Likes
4

There isn’t a specific reply to this answer, and to each their own methods, personally i’m a practical learner and I learned by coding stuff and breaking things along the way and fixing them.
If I have to say you should find something that teaches you a bit of theory to understand what is going on or at least explain what they are coding, and practice along the way with them.
I have found that DataCamp was really helpful in the earlier days of starting my journey, I also don’t recommend watching lots of videos without practicing since you tend to forget.
I have compiled a list websites where you can start learning in here sites 4 e-learning - Google Sheets feel free to add any comments that you like to add.

If I may add my personal opinion on the matter I would say these are the most useful tips I would give:

  • start coding ASAP
  • build your own projects/interfaces along the way, would be fun to build an interface using gradio or any other library to intereact with your AI model
  • share your work with others, either on social media or with other friends or with the huggingface server, the validation and constructive feedback you get from others is noteworthy
  • checkout other people’s works, sometimes you learn by checking their work and maybe they have an easier way of making things work.
  • don’t be afraid to ask questions.
  • Avoid copy pasting AI codes, if you let an AI do the entire code for you then there is no point in learning this, maybe use it as a reference to check if there is an optimized or better way to make things work, but I don’t recommend copy pasting things, since I noticed that it made learners brains get lazy and dependent (noticed this in AI events dedicated for university students)

We also have a channel for I-am-learning-today in the huggingface discord server where you can share what you are learning and a friendly community that you can ask them your questions.

1 Like
5

Since not-lain has explained most of the front roads, I will explain the back roads.
Almost everything in HF is free to copy. You can use this.

First, find a space of AI you are interested in, play with it, and copy the good ones for your own use. After a little use, you will find some dissatisfaction. Then you try to modify it, searching and asking questions about how to do it. You will probably encounter errors, but you will try to solve them by searching. You will get better and better at it, but you will be biased toward your own style, so you will read manuals and instructional websites. It is also good to learn from someone.

Once you are able to code to some extent, you should look for what you want to do at that stage.

1 Like
6

@BluePyTheDeer251

This article is useful for a gentle theoretical orientation:
What is an Embedding Anyway? | N E R | D S (medium.com)

1 Like
7

Hey! You can start off with the 365 Data Science ML course—they currently have open access for 21 days, which could be a great way to dive in and get hands-on experience. I cannot provide you a link as they’re going to delete my post but you can find it on the 365 Data Science website. Happy learning

1 Like
8

Thanks! I will use the course once I finish my PyTorch one.

1 Like