This project aims to provide a Prolog specification of individual TensorFlow layers. Most layers are specified assuming statically given weights. Taking these weights as input parameter enables a deterministic computation of outputs for given inputs.
| Layer | Comment |
|---|---|
| Add | |
| Average | |
| AveragePooling1D | |
| AveragePooling2D | |
| AveragePooling3D | |
| Concatenate | |
| Conv1D | |
| Conv2D | |
| Conv3D | |
| Conv2DTranspose | |
| Conv3DTranspose | |
| Cropping1D | |
| Cropping2D | |
| Cropping3D | |
| Dense | |
| DepthwiseConv2D | |
| Dot | |
| Embedding | |
| ELU | |
| Flatten | |
| GlobalAveragePooling1D | |
| GlobalAveragePooling2D | |
| GlobalAveragePooling3D | |
| GlobalMaxPool1D | |
| GlobalMaxPool2D | |
| GlobalMaxPool3D | |
| GRU | |
| InputSpec | |
| LayerNormalization | |
| LeakyReLU | |
| LocallyConnected1D | |
| LocallyConnected2D | |
| LSTM | |
| Maximum | |
| MaxPool1D | |
| MaxPool2D | |
| MaxPool3D | |
| Minimum | |
| Multiply | |
| Permute | |
| PReLU | |
| ReLU | |
| RepeatVector | |
| Reshape | |
| SeparableConv1D | |
| SeparableConv2D | |
| SimpleRNN | |
| Softmax | |
| Subtract | |
| ThresholdedReLU | |
| UpSampling1D | |
| UpSampling2D | |
| UpSampling3D | |
| ZeroPadding1D | |
| ZeroPadding2D | |
| ZeroPadding3D |
| Layer | Comment |
|---|---|
| AlphaDropout | |
| BatchNormalization | |
| ConvLSTM2D | |
| Dropout | |
| GaussianDropout | |
| GaussianNoise | |
| Masking | |
| SpatialDropout1D | |
| SpatialDropout2D | |
| SpatialDropout3D | |
| TimeDistributed |