|
Data Usage tab (Training Options dialog box) |
|
Use this tab to specify sections of data for training neural networks.
Displays the total number of input/output pairs available in the data set as a guide for selecting the training and test data sections.
Training Data
Specifies the data used for training the neural network, or the set of input/output pairs that are used during training.
Specifies whether the training data is a sequential range or a random set of input/output pairs.
Range
Select this option to specify that the training data is a sequential range of input/output pairs.
Random
Select this option to specify that the training data is a random set of input/output pairs chosen from a range.
Specifies the index location where the training data section begins. Type in this box a zero-based index to specify the training start position.
 | This location refers to the index of an input/output pair in the model and not to a worksheet location. |
Specifies the size of the training data section. Type in this box a number of input/output pairs to specify the size of the training data section.
 | A good suggestion for the size of this section is between 50% to 80% of the quantity displayed in the Data Set Size box. |
Specifies the index location in the data where the training data section ends. When you select Range for the Training Data Selection Mode, this box displays the end location of the training data section as a reference; otherwise, use this box to type a zero-based index for the section's end position.
| When you select Random as the Training Data Selection Mode, the total number of input/output pairs included by the start and end positions must exceed the specified Training Size. |
Displays a visual representation of the selected training region. The 
icon indicates the start position, and the 
icon indicates the end position. You can adjust the end location by sliding 
with the mouse, or by pressing the left and right arrow keys on your keyboard when the slider is selected.
Specifies the data used for testing the neural network. This data section is presented to the neural network after every training epoch to compare the results against the stop conditions.
Refer to the Heuristic Neural Training dialog box and Manual Neural Training dialog box for additional information on stop conditions.
Specifies whether the test data is a sequential range or a random set of input/output pairs.
Range
Select this option to specify that the test data is a sequential range of input/output pairs.
Random
Select this option to specify that the test data is a random set of input/output pairs chosen from a range.
Specifies the index location where the test data section begins. Type in this box a zero-based index to specify the test start position.
| This location refers to the index of an input/output pair in the model and not to a worksheet location. |
Specifies the size of the test data section. Type in this box a number of input/output pairs to specify the size of the test data section.
| A choice for the size and position of this section depends on the nature of your model. If your model represents a time series, a sample that does not overlap the training section and represents 20% to 30% of your data is a good choice. If your model represents an approximation between some bounds, it is common to set this section to the exact size and location of the training set. It is also a good practice to reserve 10% to 20% of your model data for out-of-sample tests, where you can test the performance of the neural network with data thatt was not used for training. |
Specifies the index location in the data where the test data section ends. When you select Range for the Test Data Selection Mode, this box displays the end location of the test data section as a reference; otherwise, use this box to type a zero-based index for the section's end position.
| When you select Random as the Test Data Selection Mode, the total number of input/output pairs included by the start and end positions must exceed the specified Test Size. |
Adjusts the visual representation of the selected region. The icon indicates the start position, and the icon indicates the end position of the selected region. You can adjust the end location by sliding with the mouse, or by pressing the left and right arrow keys on your keyboard.
Sponsored
Try Predictive Systems Lab
Need Help?
Contact support