Tensorflow.js tf.losses.logLoss() Function (original) (raw)

Last Updated : 17 Jun, 2021

Tensorflow.js is an open-source library developed by Google for running machine learning models and deep learning neural networks in the browser or node environment.

The Tensorflow.js tf.losses.logLoss() function calculates the log loss between two given tensors.

Syntax:

tf.losses.logLoss (labels, predictions, weights?, epsilon?, reduction?)

Parameters:

• labels: It specifies the truth output tensor. The absolute difference is predicted based on this tensor.
• predictions: It specifies the predicted output tensor with the same dimensions as labels.
• weights: It specifies a tensor of rank either equal to that of labels so that it can be broadcastable or 0. It is an optional parameter.
• epsilon: A small constant value to avoid taking log of zero. It is an optional parameter.
• reduction: It specifies the type of reduction to the loss. It is optional.

Return Value: It returns a tf.Tensor which is calculated by logLoss() function.

Example 1: In this example we will take two 2d tensors as labels and prediction. Then we will find the log loss of these two tensors.

JavaScript `

// Importing the tensorflow.js library const tf = require("@tensorflow/tfjs");

// Defining label tensor const x_label = tf.tensor2d([ [0., 1., 0.],
[1., 0., 1.] ]);

// Defining prediction tensor const x_pred = tf.tensor2d([ [1., 1., 1.],
[0., 0., 0. ] ]);

// Calculating log loss const log_loss = tf.losses.logLoss(x_label,x_pred)

// Printing the output log_loss.print()

`

Output:

Tensor 10.745397567749023

Example 2: In this example we will log loss of two given tensors and avoid taking log of zero using a small constant value, epsilon.

JavaScript `

// Importing the tensorflow.js library const tf = require("@tensorflow/tfjs");

// Defining label tensor const x_label = tf.tensor2d([ [1, 0, 0],
[1, 1, 0] ]);

// Defining prediction tensor const x_pred = tf.tensor2d([ [1, 1, 1],
[0, 0, 0] ]);

const epsilon = 0.1;

// Calculating log loss const log_loss = tf.losses.logLoss(x_label,x_pred,epsilon)

// Printing the output log_loss.print()

`