Documentation for the simulations from Norman & O'Reilly (2003, Psych Review)

1.0 Introduction
1.1 Overview
1.2 Important notes about using PDP++
1.3 Prerequisites
1.3.1 Background reading
1.4 A note on documentation
1.5 Mailing list and blog
1.6 Links to Psych Review paper

1.0 Introduction

This is the documentation for the simulations from Norman & O'Reilly (2003, Psych Review).

VERSION 1.0.1, 10/06/03

This document is available at http://compmem.princeton.edu/resources/PSYREV.README.html, or accessible from the Computational Memory lab homepage.

This document was written by Ken Norman, with help from Greg Detre.

The simulations were packaged by Ken Norman, with help from Omar Ahmed.

For announcements and discussion, see the Norman & O'Reilly weblog.

1.1 Overview

This archive contains the project files and other ancillary files that will allow you to run the simulations described in the Norman & O'Reilly (2003) Psych Review paper, "Modeling hippocampal and neocortical contributions to recognition memory: A complementary learning systems approach".

These simulations all use the PDP++ leabra algorithm implemented by Randy O'Reilly.

To run these simulations, you need to install the leabra++ simulator program, which is part of the pdp++ software package developed by Randy O'Reilly. This software is open-source and can be downloaded for free, for the Windows, Mac OS X, and Linux operating systems.

The PDP++ software home page is:

http://psych.colorado.edu/~oreilly/PDP++/PDP++.html.

** IMPORTANT NOTE: THE SIMULATIONS INCLUDED IN THIS ARCHIVE ARE DESIGNED TO RUN WITH VERSION 3.1 OF THE PDP++ SOFTWARE. THESE PROJECT FILES MAY NOT WORK PROPERLY WITH EARLIER VERSIONS OF PDP++.

These project files were created using the Linux version of leabra++. We have only cursorily tested them on Windows and Mac OS X; you can try running the simulations on Windows and Mac OS X but we can not guarantee that they will work. If you have the option of running the simulations under Linux, we strongly recommend that you do this.

Here are some direct download links:

For MS Windows:

ftp://grey.colorado.edu/pub/oreilly/pdp++/pdp++_3.1_setup.exe

For Red Hat Linux (if you don't want to compile the program on your own):

ftp://grey.colorado.edu/pub/oreilly/pdp++/pdp++-binext-3.1-0.i386.rpm.

For other forms of Linux:

ftp://grey.colorado.edu/pub/oreilly/pdp++/pdp++_3.1_binext_LINUX.tar.gz.

• type tar -zxvf pdp++_3.1_binext_LINUX.tar.gz

For Mac OS X

ftp://grey.colorado.edu/pub/oreilly/pdp++/pdp++_3.1_osx.pkg.sit.

1.2 Important notes about using PDP++

PDP++ has trouble reading script files created in certain Windows programs (notably Notepad, Wordpad and even editors like Visual C++). To be sure that PDP++ can open your script, be sure to save it using Emacs or MS Word (as a text file, but with a .css extension). If you do write it in another program and have trouble, just copy and paste into Emacs or Word and save there. Further problems can arise if you copy and paste scripts into emails. Better to send them as (compressed/archived) text file attachments.

Also, if the default font size in the Windows version is too large for your taste, you can fix this: In the PDP root Object menu, choose 'Settings', and in the 'MS Win scale' box, increase the value from 0.85 (default) to 1.4 (higher shrinks all of the text and windows). Save config. Exit and restart PDP++.

1.3 Prerequisites

We are committed to giving other people the opportunity to play with and build upon our models. We hope that, by following the instructions in the README files for the individual simulations, you will be able to get a feel for the models. If you want to modify the simulations in any way, or if you want to get a deep understanding of how the models work, you will need to dig into the background reading list below.

If you have any questions about the simulations, you should email Ken Norman. For most questions, we will end up referring you to the relevant part(s) of the background materials. If your question is not answered by the background materials, or if it comes up repeatedly, we will post an answer on the weblog.

1.3.1 Background reading

1.3.1 Background reading

• The Norman & O'Reilly (2003) Psych Review paper.

• Chapters 1, 2, 3, 4, 5, 6, 7, and 9 of O'Reilly & Munakata, (2000). Computational explorations in cognitive neuroscience: Understanding the mind by simulating the brain. Cambridge, MA: MIT Press. We also recommend that, in addition to reading the text, you try running the simulation exercises that are described in the text. The project and script files needed to run these exercises can be downloaded from:

  http://psych.colorado.edu/~oreilly/cecn_download.html. The appendices of this book are also useful to look at, although they are somewhat outdated (they refer to an earlier version of the leabra++ software that lacks some of the nice features of the current version).

• The PDP++ software, user's manual, downloadable from:

  http://www.cnbc.cmu.edu/Resources/PDP++//manual/pdp-user_1.html.

  Here are the most important sections:

  Section 2 is useful if you want to install the programmers version of leabra (i.e., the version that you compile yourself). If you want to install the end user's version, you can skip section 2 and follow the directions on the download page. If you haven't compiled UNIX programs before, then I strongly suggest NOT downloading the programmers version.

  For basic information on using the software, read the following:

  • Section 3 (conceptual overview)

  • Section 9 (projects) -- section 9.2 (project viewer) is especially useful

  • Section 10 (networks) -- sections 10.6, 10.7, and 10.8, dealing with use of the network viewer, are especially useful

  • Section 11 (environments) -- sections 11.4, 11.5, and 11.6, dealing with use of the environment viewer, are especially useful

  • Section 12 (processes) should be (at the very least) skimmed

  • Section 13 (logs) should be (at the very least) skimmed

  • Section 17 gives details on the Leabra algorithm

  Once you start to get comfortable with the simulator:

  • Section 5 (how-to guide) answers frequently asked questions about simulator details

  • Read sections 7 and 8 (guide to CSS, object basics) when you start to get curious about how scripts work.

  • Section 6 (guide to general-purpose features of the GUI) can be skipped for the first pass; look at it when you get curious about what all of the options on the various different menus mean

1.4 A note on documentation

The documentation that accompanies these simulations is a work in progress. If you have concrete suggestions regarding how to fix aspects of the documentation that are unclear, please let Ken know. We can't guarantee that we will implement these changes but we will try.

1.5 Mailing list and blog

If you would like to be contacted whenever there are major revisions to the simulations or the associated documentation files, please send your email address to Ken, with the subject line "add me to PRSIMS mailing list".

Alternatively, for announcements and discussion, see the Norman & O'Reilly weblog.

1.6 Links to Psych Review paper

Simulation 1: Pattern Separation

Figure 5 => PATSEP.CORTEX and SIMLURE.HIPPO

Simulation 2: Nature of the Underlying Distributions

Figure 6a => DISTRO.CORTEX

Figure 6b => DISTRO.HIPPO

Figure 7 => DISTRO.HIPPO

Figure 8 => BASIC.HIPPO

effects of hidden layer size in cortex: HIDSIZE.CORTEX

Simulation 3: YN Related Lure Simulations

Figure 9 => SIMLURE.CORTEX and SIMLURE.HIPPO

Figure 10 => SIMLURE.HIPPO

Simulation 4: Lure Relatedness and Test Format Interactions

Figure 11 => SIMLURE.CORTEX

Figure 12 => ENCVAR.SIMLURE.CORTEX and ENCVAR.SIMLURE.HIPPO

Simulation 5: Associative Recognition and Sensitivity to Conjunctions

Figure 15 => ASSOC.CORTEX and ASSOC.HIPPO

Figure 16 => ASSOC.SINGLECUE.HIPPO

Simulation 6: Interference and List Strength

Figure 19 => BASIC.CORTEX and BASIC.HIPPO

Figure 20 => DISTRO.HIPPO

Figure 21 => BASIC.CORTEX

Figure 22 => BASIC.CORTEX

Simulation 7: The Combined Model and the Independence Assumption

Figure 25 => INDEP.ENCVAR.COMBINED

Figure 26 => INDEP.INTERF.COMBINED

Simulation 8: Lesion Effects in the Combined Model

Figure 27 => LESION.COMBINED

Figure 28 => LESION.COMBINED

Figure 29 => LESION.COMBINED

"settle time" simulations discussed in the future directions section => LSE_SETTLE.CORTEX

NOTE: the only simulations discussed in the paper that are not included in this archive are the list-length-with-fast-weight simulations discussed in the "List Length" section. these sims are not included because they require a modification to the leabra code (and subsequent re-compiling).

Table of Contents

1.0 Introduction
1.1 Overview
1.2 Important notes about using PDP++
1.3 Prerequisites

1.3.1 Background reading

1.4 A note on documentation
1.5 Mailing list and blog
1.6 Links to Psych Review paper

About this document

• 1. Section One
• 1.1 Subsection One-One
• ...
• 1.2 Subsection One-Two
• 1.2.1 Subsubsection One-Two-One
• 1.2.2 Subsubsection One-Two-Two
• 1.2.3 Subsubsection One-Two-Three     <== Current Position
• 1.2.4 Subsubsection One-Two-Four
• 1.3 Subsection One-Three
• ...
• 1.4 Subsection One-Four