Essential Cheat Sheets for Machine Learning and Deep Learning Engineers | Hacker News Reader

Essential Cheat Sheets for Machine Learning and Deep Learning Engineers

startupsventurecapital.com

43 pointsblacksmythe13 comments

Essential to create website traffic, maybe, but not for the engineers. These things have next to no additional value over official documentation, and are of lower quality than the official docs - like the "numpy cheat sheet" telling people that "np.float32" is a standard double-precision floating point dtype ... really?

Generally cheat sheets are something you print out and put on your desk as opposed to having to search for official documentation. Because of the maximum size requirement, short and snappy references to functions / arguments for commonly used parts of the libraries are essential.

These cheat sheets are not to replace official documentation, these are to help. And we are not having a competition here, if these are useful for you use them otherwise don't.

Anyone know a good place to get started with Machine Learning?

Elements of statistical learning book

Its companion book is also very nice (and way easier for a beginner): An Introduction to Statistical Learning with Applications in R. It is one of the only textbooks I have read from cover to cover in my life.

Thanks for the suggestion that was the book I wanted to initially suggest I guess I remembered the names incorrectly

Depends on what your goals are. I do very small projects on the side just for my benefit and I've found Siraj Raval's YouTube channel to be good for digging in without much of a knowledge barrier, as well as r/learnmachinelearning. This isn't sufficient to break into a ML job obviously, but it's helped me understand some of the basic contexts of the ML world

+1 for Siraj Raval's YouTube channel. Videos are quick, entertaining and give a good overview of a lot of ai/ml concepts. Some of the content is a little over the top (i.e. memes & breaking out into song), but overall really well put together.

I agree that it's sometimes over the top, but I do prefer that over a presenter who's clearly qualified to speak on the topic but lacks the communication skills. The dude clearly cares about his productions, and I really appreciate that.

It's such a broad cross-disciplinary area, it's tough to define a good entry point that is reasonable for everyone. But I'd say actually start with getting a solid foundation on statistics (point estimation, and hypothesis testing).

If you're good with that, I'd learn a bit of DSP to get a feeling for how people in that world manipulate and clean digital signals. Some of that helps out later. FFT, DWT, PCA, ICA, Convolutional filtering, Kalman and PID filtering (if you're interested in online systems), things like that. They help especially in feature extraction for classification, but also for unsupervised methods.

Past that, the hierarchy of methods to be learned and in what order is funny, I'm not sure there is a current consensus on that.

The truth is that the best path kind of depends on your own goals. If you have a particular problem in your pocket that you want to solve, that actually helps a ton.

Unless you are very mathematical I will advise going through Jeremy Howard's course.[1][2]

First part is available with videos and exercises. I have seen videos of second part being available on YouTube. I hope rest of the material will be available soon.

[1] https://news.ycombinator.com/item?id=14640409 [2] http://course.fast.ai

We are making a user curated visual search engine to learn anything most optimally. Here is a mind map for Deep Learning for example :

https://learn-anything.xyz/machine-learning/deep-learning

Reply on news.ycombinator.com