pauloarantes / coursera-neural-net Goto Github PK

Assignments for Geoffrey Hinton's Neural Net Course on Coursera, translated from Matlab into Python.

• assignments 2-4 are quite different than what is presented in the course, as they were refactored into logical classifiers (adapted from the sklearn framework).
• more work could certainly be done to remove redundancy between assignments, especially between 3 and 4.
• course can be found here: https://www.coursera.org/course/neuralnets

• Implements linear Perceptron for two class problem

• Implements a basic framework for training neural nets with mini-batch gradient descent for a language model.
• Assignment covers hyperparameter search and observations through average cross entropy error.
  • i.e. number of training epochs, embedding and hidden layer size, training momentum

• Trains a simple Feedforward Neural Network with Backpropogation, for recognizing USPS handwritten digits.
• Assignment looks into efficient optimization, and into effective regularization.
• Recognizes USPS handwritten digits.

• Trains a Feedforward neural network with pretraining using Restricted Boltzman Machines (RBMs)
• The RBM is used as the visible-to-hidden layer in a network exactly like the one made in programming assignment 3.
• The RBM is trained using Contrastive Divergence gradient estimator with 1 full Gibbs update, a.k.a. CD-1.
• Recognizes USPS handwritten digits.