During my independent learning of Python and ML and my navigation of research tips, I was constantly inspired by the insights and generous shares of resources from the online community. In the same spirit, I dedicate this page to content and resources that may be beneficial or interesting to independent learners, where I will also add my own notes, visualizations and resources. Hope it helps and have fun learning!
Stanford CS229 ML intro with a decent level of maths and informative homework assignments
Machine Learning Specialization (Coursera) ML intro with less emphasis on maths but well-designed labs and useful tips on ML in operations
Deep Learning Specialization (Coursera) DL survey course with nice hands-on labs & interesting interviews with "DL heroes"
Berkeley CS285 Deep RL
Machine Learning: here (constantly updating)
The notes contain both conceptual and implementation details of several popular machine learning methods.
It synchronised the notes I took while taking several of the aforementioned courses and also included extended information.
It should be used for the purpose of independent learning or as a complement to the MOOCs. All typos are mine.
The Elements of Statistical Learning (Hastie, Tibshirani & Friedman)
Solution manual (Weatherwax & Epstein)
Introduction to Statistical Learning (Hastie, James, Tibshirani & Witten)
A Neural Network Playground (nice visualiser for simple feedforward neural networks with a handful of artificial training datasets)
Tools to visualise the architecture of different neural networks
A School for all Seasons on Trustworthy Machine Learning (curated by Reza Shokri & Nicolas Papernot)
百聞不如一見 。——《漢書·趙充國傳》
(lit. "hearing something a hundred times isn't better than seeing it once.")
Illustration of density with different covariance between x and y.
Illustration of model fit with different choice of the bandwidth parameter (tau) for the Gaussian weights.
Visualisation of the first two principal components of the MNIST handwritten digits dataset.
Implementation details: here.
Illustration of the decision boundary with difference values of the tuning parameter alpha in the cost complexity criterion (used for pruning).
(Note: Mindful of the selection bias. People in China cannot normally access Google and must resort to a VPN in order to do so. Unless sponsored by companies and institutions, setting up a functional VPN requires a basic level of tech skills, a decent level of search, trial and error, and sometimes unquenchable passion.)
