ML coding

In this repo I'd build a few examples to show how various ML algorithms works.

Classical Machine learning

Relevant algorithms:

• ✅Linear regression/logistic regression
  • Notebook
  • Another notebook with illustration on gradient descent
• ✅kNN classfication/regression
  • Notebook
• ✅k-means clustering
  • Notebook
• ✅Naive Bayes
  • Notebook
• ✅Cross Valdiation
  • Notebook
• ✅PCA
  • Notebook
• ✅AUC metric
  • Notebook
• Boosting/xgboost

Deep learning

Neural network general:

• Autograd
• ✅optimizers: SGD, Adam, RMSProp
  • Notebook
• ✅Gradient computation
• Backpropagation
• Integrated gradient feature importance
• Tensorboard usage

Recsys:

• ✅Collaborative filtering
  • Notebook
• Pytorch lightening
• TorchRec

NLP related:

• ✅TF-IDF
  • Notebook
• ✅CharRNN
  • blog post, Karpathy's repo
  • The goal is to redo the charRNN example in the book "hands on machine learning". There already provided an example using tf and keras
  • Notebook
• Transformer from scratch
• PyText
• ✅Contrastive decoding with Hugging
  • Notebook
• RAG. guide from LangChain
  • ✅Basic RAG
    • Notebook
  • Raptor

ML infra:

• Iterator for nested array.
• In memory SQL engine
• Spreadsheet with math equation

Reinforcement learning

• Use openAI Gym library to setup games
• PPO for CartPole
• DQN for CartPole

Statistics:

• ✅random sample with weights
  • There is a die with $n$ sides. The probability of each side of the die is $p_i$. Write a function to generate $m$ rolls from such die.
  • Notebook
• ✅design rand10() using rand7
  • Notebook