PAPERADAM: A METHOD FOR STOCHASTIC OPTIMIZATION

Covers: theory of Adam
Estimated time needed: 20 minutes
Questions this item adddesses:
  • Where does this optimizer come from?
How to use this item?

This is the optional module for the students who want to learn more about Adam. This is the first article that introduces​ this concept.

URL: https://arxiv.org/pdf/1412.6980.pdf
Shortlist
publicShare

Adam (adaptive moment estimation)

Amir PariziTotal time needed: ~2 hours
Learning Objectives
Learn the theory behind Adam as an optimizer and how to implement it in Python
Potential Use Cases
The algorithms leverages the power of adaptive learning rates methods to find individual learning rates for each parameter.
Target Audience
INTERMEDIATE
Go through the following annotated items in order:
ARTICLE 1. Intro to mathematical optimization
  • What is mathematical optimization?
  • Why do we need to optimize a cost function in ML algorithms?
10 minutes
ARTICLE 2. RMSprop
  • What is RMSprop?
  • How does this algorithm work?
10 minutes
VIDEO 3. Gradient Descent with Momentum
  • What is momentum in GD?
  • How does momentum help optimizer to perform faster?
10 minutes
VIDEO 4. Adam optimization algorithm
  • What is Adam optimizer?
  • What is the math behind this optimizer and how does adam work?
10 minutes
ARTICLE 5. Adam optimizer [more advanced math concepts behind this algorithm]
  • What is Adam optimizer?
25 minutes
LIBRARY 6. Implementing Adam optimizer in Python using Keras
  • How to implement Adam optimizer in Python?
20 minutes
PAPER 7. ADAM: A METHOD FOR STOCHASTIC OPTIMIZATION
  • Where does this optimizer come from?
20 minutes

Concepts Convered