Enhancing generalization of deep learning models with adaptive teaching learning optimization technique and t htr learning rate scheduler

Abstract

newline Neural Networks (NNs) are models with successive layers of neurons that have been in existence for decades. Training for these NNs can be done either in an unsupervised or supervised manner. The most frequently used machine learning technique for either shallow or deep networks is supervised learning. This technique will calculate an objective function that measures the error or distance between the actual result and the expected result. The learning procedure will involve adapting its internal parameters such that this error is minimized. The term weights is employed for these adjustable parameters. Deep Neural Networks (DNNs) are made up of millions of weights that may require to be updated during the training procedure. In deep learning algorithms, before training a model, the hyperparameters must be initialized. Such hyperparameters have direct control over the algorithmand#8223;s training behavior and enormous effect on the trained modeland#8223;s efficiency. Selecting an appropriate hyperparameter plays a crucial role in the success of the architecture of the NNs.