I want to explore ML bc It's just an area that interests me. Not necessarily looking to switch but not against the idea either. Are these 2 courses a solid start? Does order matter? I think Andrew Ng's new revised course is releasing next month so I'll wait until then. I want to learn hands on so I figured I should finish these 2 courses and then jump to Kaggle. If I like it, then I'll try to get more involved w ML at work/Meta. Any other advice? I don't know if there are any must read books. Tech-wise I'm comfortable w python and will focus on Pytorch since I'm already at Meta. TC: 300k depending on stock. YOE: 3
Udacity has some good “Nanodegree” courses as well. I would recommend: 1. AI Programming with Python 2. Introduction to TensorFlow 3. Introduction to PyTorch
Fastai is much better for your needs IMO. They start from explaining everything at high level and how to apply it and slowly dive deeper. Many courses including Andrew Ng’s start from building base and expand slowly. Also note that coursera course is being redesigned now so would be good to at least wait for a new version.
Fast.ai is mostly applied stuff, don’t go for it so soon. Have a look at Andrew Nag’s deep learming specialisation. That will teach you basics of Deep learning. Then you can read a PyTorch book or Fast.ai with PyTorch. Books I recommend: 1. Pattern Recognition and Machine Learning: for classical ML 2. Deep Learning by Ian GoodFellow: Deep Learning 3. Deep learning with Python by Francois Chollet: Applied DL with Tensorflow 4. Fast.ai with PyTorch by Jeremy Howard: Applied DL with tips and tricks combined. You should have solid background in ML to understand why Jeremy’s tricks work. This should be your last step.
you work at meta, just transfer to infra/backend team with ml components practice is the best way, all those courses you will forget within 1 week without doing real projects. You will also be surprised how far out of touch with reality those courses in term of industry/productionalizable ml (nit, but they also have the worst python code quality that i have ever seen) At best those courses will tell you "there is an algo X, supposedly good for Y, mkay"
Take the Andrew Ng's course and gets some hands on experience using kaggle. I suggest starting with tabular playground series since its more approachable than vision/NLP contests.
I liked Andrew Ng's course a lot but you can think of it like reading a book on CS fundamentals where you learn about graphs and implement a BFS. You will learn a ton about ML fundamentals and implement the algorithms. I followed that by reading Hands-On Machine Learning with Scikit-Learn and TensorFlow, which is a great intro to real world ML development – data pipelines, doing analysis in Jupyter Labs, using popular libraries, productionizing models, etc. With that said, ML is a fairly specialized field full of PhDs. You may find a role as a ML "practitioner" but in general, I get the impression that the ceiling is fairly low without at least a masters in a related field. If you just want to be more knowledgable and maybe get involved with a simple application of ML in your project, self-taught is fine. If you want to make it your career, go back to school.
Personal Finance
Yesterday
3067
Should I marry a lazy girl?
Tech Industry
Yesterday
1795
Absurdities of the Bay Area 😒
India
Yesterday
1136
What do vegetarian Indians eat for protein?
2024 Presidential Election
Yesterday
984
Opinion: What do you see in this photo?
India
Yesterday
704
Any Indians Think Kashmir Should be Independent?
More like, read Ian Goodfellow Deep learning and Christopher Bishop's PRML
You can watch them, but those online courses are superficial. See them and then see Andrew Ng's stanford lectures. You'll see the difference.
The Coursera course is based off his Stanford lectures