This code implements two polynomial networks for image recognition. The two codes are based on the paper of "**Π-nets: Deep Polynomial Neural Networks**" (also available here ) [1].
The two networks are implemented in both PyTorch and TensorFlow (in the folder tensorflow). Those networks aim to demonstrate the performance of the polynomial networks with minimal code examples; therefore, they are not really the state-of-the-art results on recognition. For networks that can achieve state-of-the-art results the source code of the papers can be followed, since they have more intricate implementations. For instance, for Π-nets, please check [1].
Please visit the folders of `pytorch` or `tensorflow` for implementations in PyTorch and TensorFlow respectively.
New JAX and Keras implementations for polynomial networks have been added (e.g., Minimum_example_JAX.ipynb).
The notebooks are the same as the one in the repo and contain minimum examples in PyTorch, TensorFlow, JAX and Keras respectively.
We are thankful to Yongtao for the help of converting the code to TensorFlow.
| [1] | (1, 2) https://github.com/grigorisg9gr/polynomial_nets/ |
| [2] | https://pypi.org/project/pyaml/ |