So I wanted to get into ML using Python recently and I was wondering about which ML library I should learn as a ML beginner first. I’ve been using Python for a few years now.
I’d say since you’re a beginner, it’s much better to try to implement your regression functions and any necessary helper functions (train/test split etc…) yourself in the beginning. Learn the necessary linear algebra and quadratic programming and try to implement linear regression, logistic regression and SVMs using only
numpyandcvxpy.Once you get the hang of it, you can jump straight into
sklearnand be confident that you understand sort of what those “blackboxes” really do and that will also help you a lot with troubleshooting.For neural networks and deep learning,
pytorchis imposing itself as an industry standard right now. Look up “adjoint automatic differentiation” (“backpropagation” doesn’t do it any justice aspytorchinstead implements a very general dynamic AAD) and you’ll understand the “magic” behind the gradients thatpytorchgives you. Karpathy’s YouTube tutorials are really good to get an intro to AAD/autodiff in the context of deep learning.So I should learn sklearn first before pytorch to understand the basics?
Linear and logistic regression are much easier (and less error prone) to implement from scratch than neural network training with backpropagation.
That way you can still follow the progression I suggested: implement those regressions by hand using numpy -> compare against (and appreciate) sklearn -> implement SVMs by hand using cvxpy -> appreciate sklearn again.
If you get the hang of “classical” ML, then deep learning becomes easy as it’s still machine learning, just with more complicated models and no closed-form solutions.
Aight thanks.
For more “traditional” or “statistical” modeling (not NN) 100% start with sklearn. It has a plethora of algorithms, and their docs read like a book. You can learn a whole bunch of new methods and techniques from there too. In tandum, you should familiarize yourself with matplotlib, which is the plotting library it uses under the hood (and is by far the most popular plotting library.)
For deep learning, I’d say PyTorch? Tensorflow used to be standard but its fallen out of favor compared to PyTorch. I don’t use either so I’m nit sure.
Maybe find some code to look at on the HuggingFace hub page? HuggingFace libraries or PyTorch are likely to give you really good learning opportunities and examples. Just keep an eye out for timestamps of articles or version numbers. And of course use venv/conda/… to not mess up your version when trying out different things 😉
In your opinion, is PyTorch easier than something like TF? What do you think about Keras?
I’m not personally coding with them, just often supporting people and their projects that do. Keras is also popular but I’ve at least personally seen slightly shoddier implementations with it. That could be selection bias though.
I personally think Keras has a nice and intuitive high level API for getting into nueral networks, but Pytorch is definitely the most prominent library. If your going to start somewhere you’re not going to regret learning Pytorch.
That being said, as others have mentioned, if you want to be a good data scientist or ML practioner learning the basics is never a bad idea. Sklearn is still the best library for a lot of ML tasks and is good to be familiar with.
There are a couple of good books out there that start off with the basics using numpy, pandas, Sklearn and build up to nueral networks/deep learning. I’ve use this one in the past https://www.amazon.com/Machine-Learning-PyTorch-Scikit-Learn-learning/dp/1801819319.
It’s been a while since I last looked into those.
If you aren’t looking for neural networks I found sklearn to be quite capable and easy to understand.
I also tried tensorflow and pytorch a couple times (not enough to get really proficient in them) and I think I found pytorch the hardest to wrap my head around. It’s been quite a while though so maybe it’s better to listen to others with more experience in that regard.
Thank you.
Sklearn has those built in graphs and chart displays
Sklearn for most of the data handling, pytorch for the model. They’re designed to be useable together.
What is ML?
Microwaveable Legos
Machine Learning
