Learning Progress

Learning Progress
The Learning Progress view shows how learning is progressing.  Up to 5000 graph points are recorded.  This is sufficient for over 200,000,000 learning cycles.  The graph is produced by sampling these points.  The horizontal axis is nonlinear to allow the whole learning progress to be displayed.  As more cycles are executed the graph is squashed to the left.  The scaled errors for all example rows are used.  The red line is the maximum example error, the blue line is the minimum example error and the green line is the average example error.  The orange line is the average validating error.  The errors are also shown on the Example errors view.

Learning progress

1. Learning progressThe learning progress graph shows the maximum, average and minimum training error.  The average validating error is shown if any validating examples rows are included.  The graph is squashed to the left as learning progresses so that the whole learning curve is visible.  Up to 200,000,000 learning cycles can be displayed on the graph.

Vertical scale

2. Vertical scaleThe default vertical scale is 0.0000 to 1.0000 but can be adjusted by clicking on the vertical scaling and zoom toolbar icons.  Details of the curve can be seen by scaling and zooming the graph.

Horizontal scale

3. Horizontal scaleThe horizontal scale is renumbered as learning progresses by packing to the left.  The entire learning curve is visible.


4. NetworkThe current counts of nodes and connection weights in the neural network.  This is updated if the network is reconfigured automatically or manually.

Diagnostic Array data

5. Diagnostic Array data
A copy of the data in the Diagnostic Array.  The values are updated while learning progresses.

Learning and validating

6. Learning and validatingThe target error, learning and validating progress details.  The validating and scoring results are shown when the number of validating examples is greater than 0.

Created with help of DrExplain