Skip to content
Go to file


Failed to load latest commit information.
Latest commit message
Commit time

Online Deep Learning: Learning Deep Neural Networks on the Fly

An implementation of the Hedge Backpropagation(HBP) proposed in Online Deep Learning: Learning Deep Neural Networks on the Fly

  title     = {Online Deep Learning: Learning Deep Neural Networks on the Fly},
  author    = {Doyen Sahoo and Quang Pham and Jing Lu and Steven C. H. Hoi},
  booktitle = {Proceedings of the Twenty-Seventh International Joint Conference on
               Artificial Intelligence, {IJCAI-18}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},             
  pages     = {2660--2666},
  year      = {2018},
  month     = {7},
  doi       = {10.24963/ijcai.2018/369},
  url       = {},

Link to publication

Requirements and Installation

  • Theano 0.8.2
  • Keras 1.2.1

To install HBP, you need to replace the Keras's keras/engine/ file with our modified this doesn't affect normal projects that don't use HBP. Note that as the current HBP implementation only supports Keras 1.


  • To run HBP on the sample Higgs dataset, first download the data:
wget -O data/higgs.mat
  • To train HBP, run:
cd src/hbp
python -c hb19.yaml
  • To train other baseline models, run:
cd src/baselines

Data sets

The data used in our experiments are available at

Train HBP on your own data

We provide a sample script in src/ to train HBP on a new dataset. Feel free to modify the code to suit your experiments.

Pytorch implementation

This is an independent pytorch implementation, please note that this is unofficial and not yet tested by us.

You can’t perform that action at this time.