Network responses and training sets for Dawson/Gupta paper

The directories included in the zip file provide the training files and the results for the PLOS ONE manuscript “Probability matching in perceptrons: Effects of conditional dependence and linear nonseparability” by Michael Dawson and Maya Gupta. The responses of the various networks in the two simulations that are reported are given in the two Excel spreadsheets. Each spreadsheet provides details about the training patterns, the likelihood of each pattern type being rewarded in a training set, and the responses of networks to patterns after training in separate worksheets. Comments are provided on each worksheet to explain the numbers that it contains. The other two directories provide the raw training sets used in each simulation; there are 5 for Simulation 1 and there are 20 for Simulation 2. To be run with the software they need the .net extension; however they are all simply text files. The simulations were conducted using the Rosenblatt program, which implements gradient descent training of perceptrons. Detailed descriptions of the mathematics of this program, and accounts of using the software, have been previously published (Dawson, 2004, 2005). The program itself is available as freeware from this location: http://www.bcp.psych.ualberta.ca/~mike/Software/Rosenblatt/index.html References Dawson, M. R. W. (2004). Minds And Machines: Connectionism And Psychological Modeling. Malden, MA: Blackwell Pub. Dawson, M. R. W. (2005). Connectionism : A Hands-on Approach (1st ed.). Oxford, UK ; Malden, MA: Blackwell Pub.