ADAPT Wrapper Programs
I have written several programs and scripts which wrap various ADAPT programs or implement ADAPT functionality. The general idea was to automate the running of an ADAPT study as much as possible. At the same time I have tried to implement code which allows me to avoid using the DEC workstations and the original ADAPT code. Currently I have been able to do away with the ADAPT storage mechanism. Thus, using the programs below you only need an ADAPT work area to generate descriptors. After that, you are free to take the descriptors to a Linux machine and carry on with a study and not need an ADAPT work area.

In addition some of the code (mainly descmng, which you can download here) is a harness for various subroutines which will end up in the C port of ADAPT.

Below are short descriptions along with links to man pages of various programs and driver scripts I have written to enhance tha ADAPT experience :)

The site menu also provides links which describe the workflow using these programs to generate Type I , Type II and Type III models.

  • descmng - Descriptor Manager. Allows you to store all your descriptors (no limits!), add descriptors as they come, delete them. And not only this! It also replaces dscreen, ira, mlra and lrpred. It also calculates diversity indices. And of course it can output descriptors in annlin, qnetin or pnn formats.

  • type1.py - Automates the generation of Type I models. Automatically obtains a vali value and runs comittees of annlin runs.

  • type2.py - Automates the generation of Type II models. Automatically searches multiple architectures and runs committes for each architecture, averaging the RMS errors.

  • type3.py - Automates the generation of Type III models. You need to specify the architectures,best costs, gendes.in & tsets.in and it will generate the input file for gendes and run the program. Beware it can take a long time for multiple architectures.

  • type3p.py - Similar to type3.py except that it uses Press validation and thus does not require user specified training, cv and prediction sets.