https://github.com/cyberzhg/keras-succ-reg-wrapper

A wrapper that slows down the updates of trainable weights.
https://github.com/cyberzhg/keras-succ-reg-wrapper

keras regularization wrapper

Last synced: about 1 month ago
JSON representation

A wrapper that slows down the updates of trainable weights.

• Host: GitHub
• URL: https://github.com/cyberzhg/keras-succ-reg-wrapper
• Owner: CyberZHG
• License: mit
• Archived: true
• Created: 2018-12-09T08:51:51.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2019-02-01T03:34:21.000Z (almost 6 years ago)
• Last Synced: 2024-09-30T16:08:11.779Z (about 1 month ago)
• Topics: keras, regularization, wrapper
• Language: Python
• Homepage: https://pypi.org/project/keras-succ-reg-wrapper/
• Size: 6.84 KB
• Stars: 1
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Keras Successive Regularization Wrapper

[![Travis](https://travis-ci.org/CyberZHG/keras-succ-reg-wrapper.svg)](https://travis-ci.org/CyberZHG/keras-succ-reg-wrapper)

[![Coverage](https://coveralls.io/repos/github/CyberZHG/keras-succ-reg-wrapper/badge.svg?branch=master)](https://coveralls.io/github/CyberZHG/keras-succ-reg-wrapper)

A wrapper that slows down the updates of trainable weights.

![](https://user-images.githubusercontent.com/853842/50722430-dce6c580-1109-11e9-834d-7dc92b9221db.png)

## Install

```bash

pip install keras-succ-reg-wrapper

```

## Usage

```python

import keras

from keras_succ_reg_wrapper import SuccReg

input_layer = keras.layers.Input(shape=(1,), name='Input')

dense_layer = SuccReg(

    layer=keras.layers.Dense(units=1, name='Dense'),

    regularizer=keras.regularizers.L1L2(l2=1e-3),  # Any regularizer

    name='Output',

)(input_layer)

model = keras.models.Model(inputs=input_layer, outputs=dense_layer)

model.compile(optimizer='adam', loss='mse')

model.summary()

```