Welcome to SGDOptim

SGDOptim is a Julia package for Stochastic Gradient Descent (SGD), which has become increasingly popular in solving machine learning problems, especially in the context where large-scale datasets are involved.

This package provides types and functions for users to construct (regularized) empirical risk minization problems, and use SGD or its variants to solve the problem. Specifically, the package is comprised of the following modules:

Contents:

• Construction of Optimization Problems
  • Predictors
  • Loss Functions
  • Regularizers
• Sample Streams
  • Wrap Arrays into Data Streams
• Optimization Algorithms
• Callbacks