AX Tutorial v5.0.2.0

From emergent
Jump to: navigation, search

This tutorial shows you how to construct a simple model of an actual psychological task (the CPT-AX task), from start to finish. This one tutorial touches on all major aspects of the system. It is available as an interactive emergent project in demo/leabra/ax_tutorial.proj (just run emergent, open the project and start following the instructions that appear), and in this wiki page.

The text from here is largely just copied from the ax_tutorial project, and may refer to the presence of the documentation within the project itself, and thus it probably makes the most sense to follow these directions after loading that ax_tutorial project in any case, even if you're mostly reading along on the wiki here... (NOTE: Many links in this tutorial refer to objects within emergent rather than wiki pages!)

Building a Complete Model

This project provides step-by-step directions for constructing a working neural network simulation from the ground up, including programming a simple psychological task (target detection), which we extend through several stages to ultimately simulate the more complex CPT-AX task used in working memory studies.

To return to this document at any time, just hit the ProjectDocs tab at the top of this middle panel where you are now reading.

Some basic terminology:

  • Left Browser panel is the left portion of the window with a "tree" of objects in the simulation (including the network, and the input/output data, etc).
  • Middle Editor panel is where you are currently reading -- it can display different things depending on the selected tabs at the top, and what is currently selected in the left browser panel. The left-most tab usually shows what is selected in the browser, and the other tabs with "pins" down are locked in place and contain this document and the Wizard, which we will be making heavy use of. The right-most tab represents the configuration information for the 3D display shown in the right-most view panel (which is now called "Frame1" and is empty).
  • Right Viewer panel shows 3d displays of various simulation objects, incuding the network, input/output patterns, and graphs of results, etc.

Basic Task: Target Detection

The basic task we'll simulate involves presenting letters one at at a time to the network, and having it identify "targets" from "non-targets". The targets are the letters 'A' and 'X', and the non-targets are 'B', 'C', 'Y', and 'Z'.

The network will have six input units representing each of these letters, and two output units, one for "target" and the other for "non-target". It will have a hidden layer to learn the mapping (though this task is initially so trivial that it doesn't even require a hidden layer -- we'll make it harder later).


Here are the steps we'll go through, organized as separate document chapters (which live under the docs section of the browser, as does this document):

  1. AXTut BuildNet v5.0.2.0 -- building the network
  2. AXTut InputData v5.0.2.0 -- make basic input patterns (data) to present to the network
  3. AXTut Programs v5.0.2.0 -- creating and controlling the programs that perform the simulation
  4. AXTut OutputData v5.0.2.0 -- monitoring and analyzing the performance of the model
  5. AXTut TaskProgram v5.0.2.0 -- writing a program to construct the task input patterns, including more complex tasks.
  6. Extras: elaborations that go all the way to the full CPT-AX task
    1. AXTut CPTAX_Program v5.0.2.0 -- extend our basic program to the full CPT-AX task
    2. AXTut PfcBg v5.0.2.0 -- adding a prefrontal cortex/basal ganglia to the model to handle the full CPT-AX task.

If You Get Off Track..

In the same directory where you loaded this project is a ax_tutorial_final.proj project file, which has the full project that will result from following these directions (not the extras). You can load this project and refer to it to see what things are supposed to look like.

Let's begin...

Go to AXTut BuildNet v5.0.2.0.