Slotting Learning Rate in Deep Neural Networks to Build Stronger Models

Sharma D.K.; Singh B.; Anam M.; Villalba-Condori K.O.; Gupta A.K.; Ali G.K. · Proceedings - 2nd International Conference on Smart Electronics and Communication, ICOSEC 2021 (2021) · DOI: 10.1109/icosec51865.2021.9591733

Resumen

In recent years, deep neural networks have made substantial progress in object recognition. However, one issue with deep learning is that it is currently unclear which proposed framework is exaggerated for a specific hitch. As a result, distinct dispositions are attempt before one that produces satisfactory results is discovered. This paper described a distributed supervised learning method for finding the best network architecture by modifying specifications for a perceived task dynamically. In the case of the MNIST information gathering, it is shown that asynchronous supervised learning can agree on a solution space. Setting several hyperparameters can be time-consuming when constructing neural networks. In this post, we'll provide you with some tips and instructions for better organizing your hyperparameter tuning process, which should help you find a good setting for the hyperparameters much faster.

Instituciones: Universidad Católica de Santa María