Artificial Intelligence for Big Data
by Anand Deshpande, Manish Kumar, Albenzo Coletta, Giancarlo Zaccone
Number of hidden units
The performance of the deep neural network can be tweaked by selecting and changing the number of hidden units, nh, in each of the layers. As a general guideline, it is recommended to select a larger-than-required nh value initially. This ensures enough generalization for the network. However, the higher the value of nh, the greater the computational requirement for training the deep neural network. This hyperparameter can also be tuned at the level of a layer. Each individual layer can have a different and optimal value for nh based on the results from multiple iterations on the test data. In such cases, the first layer that is connected to the input layer is recommended to be overcomplete (having more nodes than the ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access