nuget
api documents
A network class holding the layers and some helper functions for training and validation.
Convolutional neural network (CNN) is a regularized type of feed-forward neural network that learns feature engineering by itself via filters (or kernel) optimization. Vanishing gradients and exploding gradients, seen during backpropagation in earlier neural networks, are prevented by using regularized weights over fewer connections.
@author Daniel Persson (mailto.woden@gmail.com) and s.chekanov
00 Remarks
01 Syntax
02 Methods
| Name | Overloads | Summary |
|---|---|---|
| take | 1 | a helper function for VAE method implements |
| forward | 1 | Forward prop the network. |
| backward | 2 | Backprop: compute gradients wrt all parameters |
| predict | 1 | |
| SetThreads | 1 | |
| GetThreads | 1 | |
| ToString | 1 |
03 Properties
| Name | Overloads | Summary |
|---|---|---|
| LayerNum | 1 | get number of layers |
| input | 1 | get the input layer of the network |
| output | 1 | get the output layer of the network |
| BackPropagationResult | 1 | Accumulate parameters and gradients for the entire network |
| Prediction | 1 | This is a convenience function for returning the argmax prediction, assuming the last layer of the net is a softmax |
| Layer | 1 |
04 Members
Int32)a helper function for VAE method implements
| Name | Type | Description |
|---|---|---|
n | Int32 | - |
the last layer is the embedding layer for make outputs
PerformanceCounter)Forward prop the network. The trainer class passes is_training = true, but when this function is called from outside (not from the trainer), it defaults to prediction mode
| Name | Type | Description |
|---|---|---|
db | DataBlock | - |
training | PerformanceCounter | - |
Double(), PerformanceCounter)Backprop: compute gradients wrt all parameters
Int32, PerformanceCounter)Backprop: compute gradients wrt all parameters
get number of layers
get the input layer of the network
get the output layer of the network
Accumulate parameters and gradients for the entire network
This is a convenience function for returning the argmax prediction, assuming the last layer of the net is a softmax
Int32)