Protopipe

• Home
• Documentation index
• Index of cards for building neural networks

Index of cards for building neural networks

The Create neural network card offers the possibility of opening an editor to create neural network structures by connecting a few cards.

The backend for building and training neural networks is based on Keras, and almost every layer, constraint, initializer and regularizer has its corresponding card.

Note: the current version of Protopipe does not yet support the parameterization of neural networks, but this feature will be available in the future.

• Constraints
  • Max norm constraint
  • Min max norm constraint
  • Non neg constraint
  • Radial constraint constraint
  • Unit norm constraint
• Initializers
  • Constant initializer
  • Glorot normal initializer
  • Glorot uniform initializer
  • Identity initializer
  • Ones initializer
  • Orthogonal initializer
  • Random normal initializer
  • Random uniform initializer
  • Truncated normal initializer
  • Variance scaling initializer
  • Zeros initializer
• Layer wrappers
  • Bidirectional layer wrapper
  • Time distributed layer wrapper
• Layers
  • Activation layers
    • ELU layer
    • Leaky ReLU layer
    • Parametric ReLU layer
    • ReLU layer
    • Softmax layer
    • Thresholded ReLU layer
  • Attention layers
    • Additive attention layer
    • Attention layer
  • Convolution layers
    • Convolution 1D layer
    • Convolution 2D layer
    • Convolution 2D transpose layer
    • Convolution 3D layer
    • Convolution 3D transpose layer
    • Depthwise convolution 2D layer
    • Separable convolution 1D layer
    • Separable convolution 2D layer
  • Locally connected layers
    • Locally connected 1D layer
    • Locally connected 2D layer
  • Merge layers
    • Add layer (2 inputs)
    • Average layer (2 inputs)
    • Concatenate layer (2 inputs)
    • Dot layer (2 inputs)
    • Maximum layer (2 inputs)
    • Minimum layer (2 inputs)
    • Multiply layer (2 inputs)
    • Subtract layer (2 inputs)
  • Normalization layers
    • Batch normalization layer
    • Layer normalization layer
  • Pooling layers
    • Average pooling 1D layer
    • Average pooling 2D layer
    • Average pooling 3D layer
    • Global average pooling 1D layer
    • Global average pooling 2D layer
    • Global average pooling 3D layer
    • Global max pooling 1D layer
    • Global max pooling 2D layer
    • Global max pooling 3D layer
    • Max pooling 1D layer
    • Max pooling 2D layer
    • Max pooling 3D layer
  • Regularization layers
    • Activity regularization layer
    • Alpha dropout layer
    • Dropout layer
    • Gaussian dropout layer
    • Gaussian noise layer
    • Spatial dropout 1D layer
    • Spatial dropout 2D layer
    • Spatial dropout 3D layer
  • Reshaping layers
    • Cropping 1D layer
    • Cropping 2D layer
    • Cropping 3D layer
    • Flatten layer
    • Permute layer
    • Repeat vector layer
    • Reshape layer
    • Up sampling 1D layer
    • Up sampling 2D layer
    • Up sampling 3D layer
    • Zero padding 1D layer
    • Zero padding 2D layer
    • Zero padding 3D layer
  • Activation layer
  • Dense layer
  • Embedding layer
  • Masking layer
• Regularizers
  • L1 regularizer
  • L1 L2 regularizer
  • L2 regularizer
• Input
• Output