Training Analysis Page: Training Optimization

The Modify Field Dialog allows you to analyze and modify the properties of a field. The Training Analysis page allows you to view an analysis of the prediction results versus the desired values.

&  For help with predictions, see Predicting and Modeling Financial Data.

Training Optimization Sub-Page Data

This sub-page displays a graph of the fitness for each generation of the genetic optimization. This graph is the same graph that appears on the Status Dialog during optimization.

Ä    Note: The lower limit of this chart is limited to a specific value to ensure that the most relevant values are displayed. Optimization fitness values can be penalized if they do not meet certain criteria, resulting in large negative values which can distort the chart making it hard to differentiate relevant values.

In addition to the optimization curve, the following statistics are displayed.

p    Total Time

This value indicates the total amount of time that was taken in this genetic phase, including all passes. The maximum amount of time can be adjusted on the Modify Genetic Algorithm Settings Dialog.

p    Generation Yielding Best

This value indicates which generation produced the best fitness. If cross validation was being used to stop the optimization, the generation with the best cross validation is indicated.

p    Total Generations

This value indicates the number of generations that were run in the genetic optimization phase. The maximum number of generations can be adjusted on the Modify Genetic Algorithm Settings Dialog.

This sub-page is available if genetic optimization has previously been used to optimize this field. It continues to be available even if the model is trained afterwards without genetic optimization. If multiple genetic optimizations are performed on one field, only the statistics from the last optimization are displayed.

Adjusting Chart Ranges

The generation and value axes can be scaled to display specific portions of the data. The options for doing this are similar to those associated with the Chart View. For more information, see Adjusting Chart Ranges.

Training Optimization Sub-Page Analysis

The analysis presented on this page detects common characteristics of learning curves and recommends actions, if appropriate. Some common results include:

·      The best improvement on the learning was found in generation #n.

This is the normal analysis condition. The last improvement on the best cost occurred at the reported epoch.

·      The best improvement on the learning was found close to the end of optimization. Continued optimization may produce better results.

The last improvement on the best cost occurred within a few generations of the end of optimization. Typically, this indicates that better results can be achieved by performing more optimization. Optimization can be resumed on the next training by selecting Re-optimize neural network inputs and settings on save on the Training Settings page.

·      No improvement on the signal/learning was found during optimization.

This message appears when the best cost occurred in the initial generation.

If a problem occurred during the training or calculation phase, the analysis will be replaced with a description of the error. A summary of these error messages is displayed on the help for the Modify Field Dialog: Training Analysis page.

How Did I Get Here?

This is a sub-page of the Modify Field Dialog: Training Analysis page.