Optimization Loop

This is a general optimization step that optimizes the value of some parameters with respect to a cost function. It can optimize any kind of parameter as long as a cost function is provided. However, it is often used to optimize data mining model parameters.

When we train a model we need to make some choices on its parameter values, for example the number of epochs, size, or similar untrainable parameters. For these choices we rely on our experience. While this might be sufficient for some parameters, it can be less useful for others. In this last case, we make a choice, but we cannot be completely sure that it is the best choice.

That is when the optimization loop comes in handy here as it refines this initial choice. Instead of choosing one single value only, we can set a range of values and eventually assign the best one in relation to the defined cost function through the optimization loop.

The workflow shown in this video can be found on the EXAMPLES server under 04_Analytics/11_Optimization/06_Parameter_Optimization_two_examples04_Analytics/11_Optimization/06_Parameter_Optimization_two_examples^*.