Crate nncombinator

nncombinator is a neural network library that allows type-safe implementation.

Modules

• activation
  Activation Function Implementation
• arr
  Array-related data structures such as fixed-length arrays
• collection
  Definition and implementation of various collection typess
• computational_graph
  Computational graph implementation
• cuda
  Function to wrap and handle cuda kernel
• device
  Computational processes used in the implementation of neural networks
• error
  Definition of various errors
• layer
  The various layers that make up a neural network and the traits they implement
• list
  Implementation of a unidirectional list to be used for memory management in a memory pool
• lossfunction
  Implementing the loss function of a neural network
• mem
  Functions related to cuda memory
• ope
  Definition and implementation of various operations on Units
• optimizer
  Definition and implementation of optimizers to be used during training
• persistence
  Implementation on persistence of neural network models

Macros

• derive_arithmetic
  Macros for automatic derivation of the implementation of the four arithmetic operations
• derive_arr_like_arithmetic
  Macro for automatic derivation of the implementation of the four arithmetic operations of Arr,ArrView.ss

Structs

• Cons
  Stack containing elements
• Nil
  Empty stack, containing no elements

Traits

• Stack
  Trait that defines a stack to store the results computed by forward propagation when training a neural network.