Practical approach to implementing neural net architectures
While the deep neural networks are good at generalizing the training data with multi-layered iteratively-generated models, the practical application of these algorithms and theory requires careful consideration of various approaches. This section introduces general guiding principles for using the deep neural networks in practical scenarios. At a high level, we can follow a cyclic process for deployment and the use of deep neural networks, as depicted in this diagram:
We explain the preceding diagram as follows:
- Define and realign the goals: This is applicable not only to the deep neural networks but in general use of the machine learning algorithms. The use-case-specific goals related to the choice error metric and threshold target value for the metric need to be set as the first step. The goal around the error metric defines the actions in the subsequent stages of architectural design and various design choices. It is unrealistic...
