ML-ALGO

Implementation of most popular ML algorithms

• Neural networks

  Implementation of Multilayer perceptron. Used naive Stochastic Gradient Descent to minimize Loss function.

  Also, you can check your progress via progress bar (implemented via tqdm)

  

  Don't support GPU and parallel evaluations.

  List of available SGD optimizers:

  • Naive SGD
  • Momentum
  • Adam
  • RMSProp
  • AdaDelta
  • AdaGrad

  Todo in the future:

  • Batch normalization layer
  • NAG optimizer
  • Nadam optimizer
  • MiniBatch SGD
  • CNN or RNN

  Other optimizations overviewed here

  Main classes:

  • Network class
  • Loss functions
  • Activation functions

• Recommendation systems

  Implementation of Matrix factorizations algorithms.

  • Matrix factorizations
    • SVD
    • ALS
    • BPR
    • WARP