analyzer - Documentation

Analyzer module.

Source:

normal.js, line 8

Methods

(inner) analyzeConv(input, filter, strides, padding, options) → {Object}

Source:

normal.js, line 28

Calculate statistics for convolution layer

Example

let input = [5, 5, 3];
let filter = [2, 2, 8];
let strides = [1, 1];
let padding = [0, 0];
let stats = analyzer.analyzeConv(input, filter, strides, padding);

Parameters:

Name	Type	Description
`input`	Array	2 or 3 dimension array ([height, width] or [height, width, channel])
`filter`	Array	3 dimension array ([height, width, out-channel])
`strides`	Array	2 dimension array ([height, width])
`padding`	Array	2 or 4 dimension array ([right, down] or [up, right, down, left])
`options`	Object

Returns:

statistics of the convolution layer

Type: Object

(inner) analyzeFC(input, hidden) → {Object}

Source:

normal.js, line 109

Calculate statistics for fully-connected layer.

Parameters:

Name	Type	Description
`input`	Number	number of neurons in input layer.
`hidden`	Number	number of neurons in hidden layer.

Returns:

statistics

Type: Object

(inner) analyzeFlatten(input, hidden) → {Object}

Source:

normal.js, line 126

Calculate statistics for flatten layer.

Parameters:

Name	Type	Description
`input`	Array	number of neurons in input layer.
`hidden`	Number	number of neurons in hidden layer.

Returns:

statistics

Type: Object

(inner) analyzeNetwork(networkJson, input, options) → {Object}

Source:

normal.js, line 149

Calculate statistics of a network

Parameters:

Name	Type	Description
`networkJson`	Object	network's architecture in json format
`input`	Array	2 or 3 dimension array ([height, width] or [height, width, channel])
`options`	Object

Returns:

statistics

Type: Object

(inner) analyzePool(input, filter, strides, padding, options) → {Object}

Source:

normal.js, line 73

Calculate statistics for pooling layer

Parameters:

Name	Type	Description
`input`	Array	2 or 3 dimension array ([height, width] or [height, width, channel])
`filter`	Array	2 dimension array ([height, width])
`strides`	Array	2 dimension array ([height, width])
`padding`	String	The padding algorithm. Either "VALID" or "SAME".
`options`	Object

Returns:

statistics of the pooling layer

Type: Object