neupy 0.2.0b1

NeuPy is the Artificial Neural Network library implemented in Python.

NeuPy v0.2.0 (beta)

NeuPy is a Python library for Artificial Neural Networks.

Installation

$ pip install neupy

Links

• Documentation

• Issues

• Tutorials

• Available algorithms

Tutorials

• Password recovery

• Discrete Hopfield Network

• Boston house-prices dataset

• Visualize Backpropagation Algorithms

Dependencies

• Python 2.7, 3.4

• Theano >= 0.7.0

• NumPy >= 1.9.0

• SciPy >= 0.14.0

• Matplotlib >= 1.4.0

Next steps

• Adding convolutional neural network layers (Issue #56)

• Adding reccurent neural network layers (Issue #57)

• Bug fixing and version stabilization (Known bugs)

• Speeding up algorithms

• Adding more algorithms

Algorithms

• Algorithms that use Backpropagation training approach

  • Classic Gradient Descent

  • Mini-batch Gradient Descent

  • Conjugate Gradient

    • Fletcher-Reeves

    • Polak-Ribiere

    • Hestenes-Stiefel

    • Conjugate Descent

    • Liu-Storey

    • Dai-Yuan

  • quasi-Newton with Wolfe search

    • BFGS

    • DFP

    • PSB

    • SR1

  • Levenberg-Marquardt

  • Hessian (Newton’s method)

  • Hessian diagonal

  • Momentum

  • Nesterov Momentum

  • RPROP

  • iRPROP+

  • QuickProp

  • Adadelta

  • Adagrad

  • RMSProp

  • Adam

  • AdaMax

• Algorithms that penalize weights

  • Weight Decay

  • Weight Elimination

• Algorithms that update learning rate

  • Adaptive Learning Rate

  • Error difference Update

  • Linear search using Golden Search or Brent

  • Search than converge

  • Simple Step Minimization

• Ensembles

  • Mixture of Experts

  • Dynamically Averaged Network (DAN)

• Neural Networks with Radial Basis Functions (RBFN)

  • Generalized Regression Neural Network (GRNN)

  • Probabilistic Neural Network (PNN)

  • Radial basis function K-means

• Autoasociative Memory

  • Discrete BAM Network

  • CMAC Network

  • Discrete Hopfield Network

• Competitive Networks

  • Adaptive Resonance Theory (ART1) Network

  • Self-Organizing Feature Map (SOFM or SOM)

• Linear networks

  • Perceptron

  • LMS Network

  • Modified Relaxation Network

• Associative

  • OJA

  • Kohonen

  • Instar

  • Hebb