- Demo
- Initial Homomorphic Encryption Support (syft/he)
- Basic Linear Model (syft/nn/linear.py)
- Initial Benchmark Testing Suite
- Proof of Concept
- Homomorphic Encryption (syft/he)
- Add Generic Tensor as Basic Type (ISSUE 10)
- Add Additive/Multiplicative Depth Tracking (ISSUE 11)
- Abstract Fixed Point Precision (ISSUE 12)
- CPU YASHE Support (ISSUE #)
- Initial Wrap of cuYASHE Support (ISSUE #)
- Internal, faster "from scratch" rebuild of Paillier (ISSUE 13)
- Key-Rotation Support for Additive HE (ISSUE 14)
- Key-Rotation Support for Add+Mul HE (ISSUE #)
- Full Unit-Testing Suite for All Features (ISSUE #)
- Benchmarking Suite for All Operations & Encryptions (ISSUE #)
- Development server for Fast Decryption/Re-Encryption
- RESEARCH: whether https://eprint.iacr.org/2016/870.pdf can be used to speedup YASHE/FV bootstrapping.
- Neural Networks
- Components
- Full Unit-Testing Suite for All Components (ISSUE #)
- Develop Abstraction for Encrypted vs Unencrypted Logic in a Layer (ISSUE #)
- Layers
- Linear Layer (ISSUE #)
- Convolutional Layer (ISSUE #)
- Embedding Layer (ISSUE #)
- Hashed Embedding Layer (variable length vocab) (ISSUE #)
- Softmax Layer (ISSUE #)
- Sparsemax Layer (ISSUE #)
- Hierarchical Layer (softmax/sparsemax) (ISSUE #)
- Vanilla Recurrent Layer (ISSUE #)
- LSTM Layer (ISSUE #)
- Nonlinearities
- Add Support for ReLU (ISSUE #)
- Add Support for Sigmoid (ISSUE #)
- Add Support for Tanh (ISSUE #)
- Add Support for Hard Tanh (ISSUE #)
- Add Experimental Support for x2 and x3 (ISSUE #)
- Losses
- MSE Loss (ISSUE #)
- Cross Entropy Loss (ISSUE #)
- Pre-fab Classifiers
- Linear Classifier (ISSUE #)
- Convolutional Neural Network (ISSUE #)
- Word2vec Classifier (ISSUE #)
- FastText Classifier (ISSUE #)
- LSTM Classifier (ISSUE #)
- RL Model (ISSUE #)
- Components
- Homomorphic Encryption (syft/he)