NeuroShell 2 Batch Processor: What's Going on Under the Hood

Top  Previous  Next

You do not have to read this portion of the documentation to understand how to use the batch processor.  It is provided only to give you an understanding of how the batch processor executes internally, including file copying and renaming, when files may be deleted, and what is happening during session execution.  Also provided in this section are some do's and don'ts while using the NeuroShell 2 Batch Processor.


How Does Session Execution Work?

Unlike NeuroShell 2, the NeuroShell 2 Batch Processor allows you to use file names with different prefixes, and to store files in different directories.  In order for this to work, several actions are performed behind the scenes:


1.  The Batch Processor first checks the training files, the minimum and maximum file, the configuration file, and any apply to files to see if they have the same directory names and file name prefix as the problem directory.


Note: The problem directory is the directory that you specify when defining the problem name and description in the batch grid of the Batch Processor.


2.  If the file name has a different prefix than the problem name but is in the same directory as the problem, the file will be renamed to a file with the problem name as the prefix and its previous suffix.




Problem name is Stock in the \NSHELL2\TEST directory.

File name is NYSE.PAT in the \NSHELL2\TEST directory.


NYSE.PAT is renamed to STOCK.PAT.


3.  If the file is in a different directory than the problem, the file will be copied into the problem directory.  It will be named with the problem name as the prefix while retaining its previous suffix.




Problem name is Stock in the \NSHELL2\TEST directory.

File name is NYSE.PAT in the \DATA directory.


NYSE.PAT is copied to the \NSHELL2\TEST directory and renamed to STOCK.PAT.




A.  When copying files, the file name may or may not be changed depending upon whether or not it already contains the problem name prefix.


B.  If a file will be copied to or renamed in the problem directory and the file already exists, the existing file will be deleted permanently from the directory.



Once the Batch Processor has made sure all of the files are properly named and are in the same directory, training, network application, and the generation of a runtime definition file will proceed.  Once these actions have been completed, the Batch Processor will return any files to their original name if they were renamed.  Copied files will not be removed.


Steps 1 through 4 will be repeated for each problem in each sequence of the currently executing session.


Some Do's and Don'ts

Now that you understand the nature of how the Batch Processor accomplishes some of its tasks, let's go over a few helpful hints which will make using the Batch Processor easier and faster.


1.  Since the Batch Processor copies files which are in directories other than the problem directory, try to store all of the files for a problem in one directory to increase the speed of execution.  Renaming a file is obviously faster than copying a file, and as files grow larger in size and as you include more problems in a session, the speed increase can become noticeable.


2.  If a duplicate file is found in the problem directory during a file copy operation, that file is removed permanently.  This usually will not be a problem, but if you want to retain exact copies of your files, it is a good idea to make a backup of them prior to session execution.


3.  Try verifying your session as you go along, maybe every two or three problems.  Often it is easier to correct one or two problems than it is to correct many.