Leetcode

ML Code

Gradient Descent

  • define loss function and solve
  • if gradient exploding ➡ clipping

K-Means

Logistic Regression

  • how to take derivatives?
  • how to convert to matrix multiplication?

Linear Regression

  • closed form?
  • using gradient descent

Attention

  • self attention
  • cross attention

other ML algos from scratch

Project / Experience Specific

General ML Knowledge

Metrics

  • accuracy
  • precision
  • recall
  • F1
  • ROC-AUC

naive bayes

  • how to derive from original form to naive form?

SVM

  • what’s supporting vector?
  • what’s advantage
  • why it’s not easy to overfitting?
    • hinge loss

Logistic Regression

  • how to get its loss function

Tree

  • ID3
  • C4.5
  • CART
  • pruning
  • random forest
  • adaboost
  • gbdt

unsupervised

  • K-Means
  • PCA

DL Knowledge

NN

CNN

RNN

LSTM

Transformer

NLP stuffs

  • like word2vec

(optional) recommendation system

  • content based
  • collaborative filtering
  • matrix factorization
  • DeepFM
  • GNN
  • lightGCN

Other Chores

Dropout

Activation Functions

  • sigmoid
  • tanh
  • relu
  • leaky relu
  • prelu
  • swish

Normalization

  • batch norm
  • layer norm

KD Divergence

Initialization

  • Xavier
  • He
  • why didn’t random nor all 0 work?

Optimizer

  • SGD
  • momentum SGD
  • adagrad
  • rmsprop
  • adam

ML Sys Design

➡ bytebytego