Liquid Carrot - activation

Activation functions

Activation functions determine what activation value neurons should get. Depending on your network's environment, choosing a suitable activation function can have a positive impact on the learning ability of the network.

Source:

methods/activation.js, line 24

See:

Example

      ;

// eg.
A.squash = methods.activation.LOGISTIC;'>Copy
      ;

// eg.
A.squash = methods.activation.LOGISTIC;'>let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.<ACTIVATION_FUNCTION>;

// eg.
A.squash = methods.activation.LOGISTIC;

Methods

(static) ABSOLUTE (x, derivateopt) open an issue

Absolute function.

Avoid using this activation function on a node with a selfconnection

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.ABSOLUTE;

Source:

methods/activation.js, line 458

(static) BENT_IDENTITY (x, derivateopt) open an issue

Bent identity function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.BENT_IDENTITY;

Source:

methods/activation.js, line 321

(static) BIPOLAR (x, derivateopt) open an issue

Bipolar function, if x > 0 then returns 1, otherwise returns -1

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.BIPOLAR;

Source:

methods/activation.js, line 355

(static) BIPOLAR_SIGMOID (x, derivateopt) open an issue

Bipolar sigmoid function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.BIPOLAR_SIGMOID;

Source:

methods/activation.js, line 388

(static) GAUSSIAN (x, derivateopt) open an issue

Guassian function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.GAUSSIAN;

Source:

methods/activation.js, line 287

(static) HARD_TANH (x, derivateopt) open an issue

Hard tanh function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.HARD_TANH;

Source:

methods/activation.js, line 422

(static) IDENTITY (x, derivateopt) open an issue

Identity function.

Returns input as output, used for memory neurons.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.IDENTITY;

Source:

methods/activation.js, line 110

(static) INVERSE (x, derivateopt) open an issue

Inverse function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.INVERSE;

Source:

methods/activation.js, line 492

(static) LOGISTIC (x, derivateopt) open an issue

Logistic function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.LOGISTIC;

Source:

methods/activation.js, line 40

(static) RELU (x, derivateopt) open an issue

ReLU function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.RELU;

Source:

methods/activation.js, line 178

(static) SELU (x, derivateopt) open an issue

Scaled exponential linear unit.

Exponential linear units try to make the mean activations closer to zero which speeds up learning. It has been shown that ELUs can obtain higher classification accuracy than ReLUs. α is a hyper-parameter here and to be tuned and the constraint is α ≥ 0(zero).

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.SELU;

Source:

methods/activation.js, line 530

See:

Self-Normalizing Neural Networks

(static) SINUSOID (x, derivateopt) open an issue

Sinusoid function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.SINUSOID;

Source:

methods/activation.js, line 253

(static) SOFTSIGN (x, derivateopt) open an issue

Softsign function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.SOFTSIGN;

Source:

methods/activation.js, line 212

(static) STEP (x, derivateopt) open an issue

Step function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.STEP;

Source:

methods/activation.js, line 144

(static) TANH (x, derivateopt) open an issue

TanH function.

Parameters:

Name	Type	Description
`x`	number \| Array.<number>	Input value(s) to activation function
`derivate`	boolean <optional>	Flag to select derivative function

Example

      Copy
      let { methods, Node } = require("@liquid-carrot/carrot");

// Changing a neuron's activation function
let A = new Node();
A.squash = methods.activation.TANH;

Source:

methods/activation.js, line 74