The Regularisers in Tensorflow.js are attached with various components of models which work with the score function to help drive trainable values, large values. The method tf.regularizers.l2 () is inherited from regularizers class. The tf.regularizers.l2() methods apply l2 regularization in penalty case of model training. This method adds a term to the loss to perform penalty for large weights.It adds Loss+=sum(l2 * x^2) loss. So in this article, we are going to see how tf.regularizers.l2() function works.
Syntax:
tf.regularizers.l2 (args);
Parameters:
- l2: The number represents the regularization rate by default it is 0.01.
Return: Regularizer
Example 1: In this example, we are going to see the standalone use of l2 Regularizer applied to the kernel weights matrix.
Javascript
// Importing the tensorflow.js library const tf = require( "@tensorflow/tfjs" ); // Define sequential model const model = tf.sequential(); // Add layer to it model.add(tf.layers.dense({ units: 32, batchInputShape:[ null ,50], kernelRegularizer:tf.regularizers.l2() })); // Model summary model.summary(); |
Output:
Layer (type) Output shape Param # ================================================================= dense_Dense1 (Dense) [null,32] 1632 ================================================================= Total params: 1632 Trainable params: 1632 Non-trainable params: 0
Example 2: In this example, we are going to see the standalone use of l2 Regularizer applied to the bias vector.
Javascript
// Importing the tensorflow.js library const tf = require( "@tensorflow/tfjs" ); // Define sequential model const model = tf.sequential(); // Add layer to it model.add(tf.layers.dense({ units: 32, batchInputShape:[ null ,50], biasRegularizer:tf.regularizers.l2() })); // Model summary model.summary(); |
Output:
Layer (type) Output shape Param # ================================================================= dense_Dense2 (Dense) [null,32] 1632 ================================================================= Total params: 1632 Trainable params: 1632 Non-trainable params: 0
References:https://js.tensorflow.org/api/latest/#regularizers.l2