MPIL version 1.0 Copyright (C) 1995 Universal Problem Solvers All Rights Reserved ***About System*** MPIL (Muli-Pass Instance-Based Learning) Version 1.0 Demo is freely available (FreeWare) to anyone and may be distributed to others, as long as it is ONLY utilized for personal or educational purposes. This Demo version restricts several parameter options (i.e., Maximum number of training patterns, inputs, outputs, etc.). A complete, unrestricted system may be purchased if so desired. For more information refer to the bottom of this document. ***Purpose*** MPIL is an instance-based learning system (instances are simply viewed as points in n-dimensional real-space with an associated neighborhood), which utilizes two models for creating neighborhoods. The first model (i.e., MPIL-1) places a single neighborhood sphere (based on Euclidean distance measure) around an instance, and is in nature similar to the nearest neighbor classifier, except that it removes redundant instances. The second model (i.e., MPIL-2) incorporates N radii (one for each input of an instance). This model also supports knowledge acquisition in the form of rule extraction. In a sense, both approaches are similar to neural networks in that they exploit a very similar parallelism. MPIL represents a good alternative in cases were large amounts of data have to be learned and provides good facilities for storage reduction. ***Highlights*** (1) Allows user to create an abstract instance representation of a training set. (2) Provides features for Saving and Loading the abstract instance representation. (3) Supports two modes for instance-based learning: MPIL-1 and MPIL-2. (4) Supplies the user with the capability to test and classify new patterns. (5) Allows batch training and testing of a data set (i.e., n-fold crossvalidation) for a user defined start partition size, end size, delta stepsize and parameter n. *** Installation *** First create directory MPIL-DEM in the root directory of the C drive. Then Place the zip file MPIL10.ZIP into C:\MPIL-DEM\. Unzip the file and preserve its subdirectory structure. Next, install the code with Windows Program Manager: (1) Select File menu option of Windows Program manager. (2) Select menu option New and check the Program Group radio button. Then select OK. (3) For the Description of the Program Group Properties enter: MPIL v. 1.0 (4) Now select OK button. (5) Next, again select the New option from the File menu of the Windows Program Manager. (6) This time the Program Item radio button should be selected. Hit OK. (7) Now enter for the selections: Description: Multi-Pass Instance-Based Learning Command Line: C:\MPIL-DEM\MPIL.EXE Working Directory: C:\MPIL-DEM (8) Once you have entered these choices hit OK button. You should now be able to view the MPIL icon within the earlier created folder MPIL v. 1.0. You are ready to get started. *** Requirements *** The version requires Windows 3.1 or Windows for Workgroups. ***IMPORTANT: Getting Started*** As a simple example you may want to try out the following learning problem: Once you have started MPIL, select Create from the Instance option of the main menu. You will be prompted to enter a training file (i.e., extension *.net). Using the directory browser switch to directory DOMAINS within the working directory C:\MPIL-DEM. Now select the file OR6.NET and open it. After a little time has passed, you will be informed that the training patterns of the 6 input OR function have been loaded. Proceed by selecting Train from the Instance option of the main menu. Training and creation of abstracted instances will initiate. Once this process is completed, a message box will appear with some pertinent information about the training cycle. For example, you will be informed about how long training lasted and what the savings ratio of training patterns is. Since the default learning mode is MPIL-1, you may consider for the next experiment, to switch to learning mode MPIL-2. To achieve this, simply select Learning Mode from the Parameters menu option. Push the MPIL-2 radio button and then hit the OK button. Next, repeat the steps of creating and training the OR6.NET data set. Once you have completed this, you can recall previously obtained results by selecting the appropriate file option from the Instance menu option (located below the double-bar). MPIL maintains up to 12 such result options. *** Documentation *** For documentation, consult the on-line help option. *** List of Files *** The MPIL 1.0 package consists of the following files: README.TXT introductory information (this file) MPIL10.EXE required executable and 7 help files (i.e., extension *.txt). DOMAINS Sub-directory: C-HEART.NET Cleveland heart disease data set ECHO.NET Echocardiogram data for heart attacks GLASS.NET Glass classification data H-HEART.NET Hungarian heart disease data set HEPATITI.NET Hepatitis data set IONO.NET Radar data of ionosphere IRIS.NET Fisher Iris Plant data LED.NET Light-emitting diode data LENSES.NET Lenses data LIVER.NET BUPA liver disorders data LUNG.NET Lung cancer diagnosis OR2.NET 2 input or-function OR3.NET 3 input or-function OR4.NET 4 input or-function OR5.NET 5 input or-function OR6.NET 6 input or-function S-HEART.NET Switzerland heart disease data SHUTTLE.NET Space shuttle maneuvering data VA-HEART.NET V.A. Medical Center California heart disease data VOTE.NET U.S. House votes WINE.NET wine classification data Note most of the data sets provided, were obtained from the UCI repository of machine learning databases. These data sets may be distributed for non-commercial use. For more information on this database contact Patrick Murphy (pmurphy@ics.uci.edu). Acknowledgment: Thanks goes out to Andras Janos who supplied the Hungarian heart disease data set, and Robert Detrano for the V.A. Medical Center heart disease data. IRIS.TES file contains iris test patterns. OR2.TES file contains or2 test patterns. OR6.TES file contains or6 test patterns. OR2.CLS file contains or2 classification patterns. ***Extended Version*** If you like the MPIL Demo software, and if you are interested in purchasing the complete system, then contact Universal Problem Solvers via electronic mail to zlxx69a@prodigy.com, or by sending postal mail to 610 South Duncan Avenue, Clearwater, FL 34616 USA. The complete version is available for US$10 + US$3 Shipping and Handling. Comments and suggestions for improvement are also welcomed. ***Other Products*** The following product is also available and a DEMO is located at the SimTel site "\win3\neurlnet\": TDL (Trans-Dimensional Learning) v. 1.01 ###