What's chagned

1. Added Embedding layer

2. Added batch size support for LSTM and GRU layers

3. Fixed a few smalls bugs regarding printing the progress of learning

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v1.2.0...v1.2.1

What's changed

1. Added GRU layer

2. LSTM layer can now update it's parameters

3. Added an option to optimizers to specify whether you want to update the biases or no

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v1.1.1...v1.2.0

What's changed

1. Made parameter initalizers their own class

2. Added 6 more parameter initalizers

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v1.1.0...v1.1.1

What's changed

1. Added MinPool for 1D and 2D

2. Added AvgPool for 1D and 2D

3. Added batch size support during training

4. Fixed a few bugs related to Convolutional and Pooling layers

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v1.0.1...v1.1.0

What's changed

1. Implemented LearningRateScheduler and CSVLogger

2. Callbacks now work in place

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v1.0.0...v1.0.1

Version 1.0.0 finally released!

What's changed:

Conv2D layers are a whole lot faster

MaxPooling layers are now working correctly

Added layers parameter to NN.init()

Fixed a lot of small bugs

Additional changes

From now on I will try to make the project backward compatible as much as I can

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v0.10.1...v1.0.0

What's changed

1. Dense layers are a lot better now

2. Added set_weights and get_weights

3. Added validation_split to NN.train()

4. A few small bug changes

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v0.10.0...v0.10.1

What's changed

1. MaxPool2D is re-re-enabled as I've fixed the backpropagation for it

2. Better ETA showcase

3. Val accuracy is now displayed properly

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v0.9.2...v0.10.0

What's Changed

Conv2D layer is now up to 6x faster

Disabled MaxPool2D layer as I found a bug in there

Full Changelog: https://github.com/MarcelWinterot/nano-keras/compare/v0.9.1...v0.9.2

What's Changed

nano-keras is now available on PyPI at https://pypi.org/project/nano-keras/

Note that there will be no backward compatibility until I release v1.0.0 so expect changes