Machine Learning Notebook

Introduction

This is my personal notebook for documenting knowledge I picked up as I progress through my career in machine learning. I like to write things down to reinforce my understanding of a topic. Although I strive to provide the best explanation, I don't do this full time. I don't recommend this notebook as a learning resource for beginners.

If you are reading this, I recommend the following resources for you. They are written by people in the research communities.

Python 2 vs Python 3

I wrote majority of the content in Python 2.7 in 2018. Now it's 2023, Python 2 has been long deprecated, I am switching to Python 3.8 with TensorFlow 2.x and PyTorch.

My current system setup

Ubuntu 20.04
Tensorflow 2.8 or PyTorch 1.13
Python 3.8.*
CUDA 11.2
cuDNN 8.4
Matplotlib 3.5.*

Some older code will be running on

Tensorflow 1.15
Python 2.7.*

PyTorch 2.0 is coming out in March 2023. I will switch to that soon.

Clustering
Simple Neural Networks
Convolutional Neural Networks
Generative Adversial Networks
Recurrent Neural Networks
Random Forest
Reinforcement Learning
Natural Language Processing
Naive Bayesian Networks
Recommender System
Transferred Learning
Machine Learning in Production

Export Notebook

Jupyter Convert

If my notebook does not contain any matplotlib.pyplot then I can export it as simple text.

jupyter nbconvert --to markdown loss_function_overview.ipynb --stdout

Otherwise, I'd need to export differently.

jupyter nbconvert --to markdown loss_function_overview.ipynb

Latex

Jupyter notebook uses single dollar sign for inline equations but GitBook uses double dollar sign for inline equations. I need a RegExp that capture and convert.

? means once or none.

\$.?\$

+ means one or more.

\$.+\$

The following will capture all $<some text>$ .

^\$.+\$$