hack your deep learning performance
Project description
Darkon: Performance hacking for your AI model
Darkon is an open source software library for improving and debugging deep learning models. People think that deep neural network is a black-box that requires only large dataset and expect learning algorithms returns well-performing models. However, trained models often fail in real world usages, and it is difficult to fix such failure due to the black-box nature of deep neural networks.
Darkon will gradually provide performance hacking features easily applicable to your existing projects based on following technologies. - Dataset inspection/filtering - Continual learning - Meta/transfer learning - Hyper parameter optimization - Network architecture search
In this first release, we provide influence function calculation feature easily applicable to any Tensorflow models. Influence score can be used for filtering bad training samples that affects test performance negatively. It can be used for prioritize potential mislabeled examples to be annotated, and debugging distribution mismatch between train and test samples.
More features will be released soon. Please keep your eyes on Darkon
Dependencies
Installation
pip install darkon
Examples / Getting Started
Documentation: [STRIKEOUT:will be soon]
Communication
Issues: report issues, bugs, and request new features
News: [STRIKEOUT:link twitter account]
Discuss: [STRIKEOUT:gitter]
Email: darkon@neosapience.com
License
Apache License v2.0
References
[1] Pang Wei Koh and Percy Liang “Understanding Black-box Predictions via Influence Functions” ICML2017]
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distributions
Built Distribution
Hashes for darkon-0.0.1-py2.py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 38211378c1a490854d719ae28cb73be3b989730b1e6d0b89ff04eb43508323f2 |
|
MD5 | 6e040f907cfb285895d99b1a6841fffc |
|
BLAKE2b-256 | ec343c8a6740c47c251b88d6f10ec80a97be1041ec3ea3c9185a628fa862ad4d |