All Versions
Latest Version
Avg Release Cycle
47 days
Latest Release
1837 days ago

Changelog History

  • v0.8.2 Changes

    April 04, 2019

    Main changes:

    • ๐Ÿ‘Œ Improved functionality for layer/network copying
    • โž• Added repeat function
    • ๐Ÿ›  Fixed output shape for the conv layers when input is unknown
    • ๐Ÿ›  Typo fixes
  • v0.8.1 Changes

    March 30, 2019

    Main changes:

    • โž• Added Tensorflow 1.13 support
    • โž• Added DropBlock layer
    • โž• Added define method for all layers
  • v0.8.0 Changes

    January 28, 2019

    Main changes:

    • ๐Ÿ†• New design for the layer graphs and layers
    • ๐Ÿ”„ Changed inline operators > (still supported) to >> and lists to| operator for parallel connections
    • Added 2 new step decay algorithms, namely exponential_decay and polynomial_decay
    • ๐Ÿ†• New regularizers
    • ๐Ÿ†• New signals functionality, can be extended with classes (signals attribute)
    • โšก๏ธ Tensorflow optimizers are used instead of self-implemented functions
    • โž• Added GroupNorm layer
    • ๐Ÿ”„ Changed behavior for the the predict methods
    • โž• Added show method to the networks
    • โž• Added plot_errors for the optimizers with better and more accurate visualizations

    Small changes:

    • โšก๏ธ The error parameters for optimizers was renamed to loss
    • ๐Ÿ”„ Changed summary format in the logs
    • Removed training_errors and validation_errors and it were replaced with new errors attribute
    • ๐Ÿ“‡ Renamed the prediction_error method to score


    • ๐Ÿ”จ Big code refactoring
    • ๐Ÿ‘ Better exceptions

    โœ‚ Removed

    • โž• Add-ons were removed
    • ๐Ÿ”ง Tuple of integer as network configuration
    • โœ‚ Removed RBFKmeans
    • Removed train_end_signal and epoch_end_signal
    • โœ‚ Removed plots.error_plot function
    • โœ‚ Removed plots.network_structure function
  • v0.7.3 Changes

    January 17, 2019
    • ๐Ÿ›  Fixed problem that occurred when GNG was loaded from pickled file. Issue #229
    • ๐Ÿ‘Œ Improved representation for the GNG graph elements
    • โž• Added extra exceptions into GNG graph methods in order to get informative fail when unexpected action was applied
  • v0.7.2 Changes

    December 13, 2018

    ๐Ÿ”„ Changes:

    • โž• Added missing dilation rates for ResNet-50 architecture
    • ๐Ÿ›  Fixed fan computation for weight initializers
    • โž• Added variable initializer in the predict method
    • ๐Ÿ”„ Change XavierNormal to HeNormal default initializer
  • v0.7.1 Changes

    December 10, 2018

    ๐Ÿ”„ Changes:

    • Speed up network initialization via lazy parameter initialization using tensorflow's tensors instead numpy's arrays
    • Global pooling layer accepts two string arguments that point to different tensorflow functions.
    • ๐Ÿ›  Fixes for the reshape layer when used with unknown input shape
    • ๐Ÿ›  Fixed cross entropy loss functions for spatial inputs
    • โœ‚ Removed input blocker during the training
    • Combined GradientDescent and MinibatchGradientDescent into one class GradientDescent
  • v0.7.0 Changes

    December 02, 2018

    ๐Ÿ”„ Changes:

    • ๐Ÿšš Backend was moved to Tensorflow
    • Pickle storage for the weights has been replaced with HDF5
    • ๐Ÿ”„ Changed order of the dimensions for the convolutional filter (channel expected to be in the last dimension)
    • ๐Ÿšš The compile method was removed
    • โž• Added wolfe search to conjugate gradient
    • ๐Ÿ›  Fixes for the training algorithms

    โœ‚ Removed:

    • Linear models
    • Quickprop training algorithm
    • Ensemble algorithms
  • v0.6.5 Changes

    October 12, 2018

    ๐Ÿ› Bugs:

    • ๐Ÿ›  Fixed iterative updates for discrete hopfield networks: #218
  • v0.6.4 Changes

    March 26, 2018

    ๐Ÿ”‹ Features:

    • โž• Added Growing Neural Gas (GNG) algorithm
    • โž• Added 2 new improvements to the base GNG algorithms

    ๐Ÿ› Bug:

    • 0๏ธโƒฃ Set up default value for epoch_time in order to prevent crashes when training was stopped on first iteration.
  • v0.6.3 Changes

    February 03, 2018

    ๐Ÿ›  Fixes:

    • ๐Ÿ’… Use old table-style, otherwise table breaks for some of the font styles and it doesn't look good in ipython notebook
    • โšก๏ธ Adam optimizer fix for GPU (#200)