This program is a Neural Net to determine which of 3 possible cultivars of wine a set of data is based on these characteristics:
- Alcohol
- Malic acid
- Ash
- Alcalinity of ash
- Magnesium
- Total phenols
- Flavanoids
- Nonflavanoid phenols
- Proanthocyanins
- Color intensity
- Hue
- OD280/OD315 of diluted wines
- Proline
The program compiles with make or make clean. You can enter the amount of epochs to loop through the input file, and the amount of lines (1-178) to loop through.
The output values are rated from -1 to 1 based which cultivar the network thinks it is.
The network uses 13 input neurons, a hidden layer of 6 neurons, and an output layer of 3 neurons. Each output neuron outputs a value from -1 to 1. The first neuron represents cultivar 1, output neuron 2 is cultivar 2, and the same for output neuron 3.
The data is normalized using a zero unit variance function: data = (data - mean_data) / standard deviation
Each neuron applies a hyperbolic tan function to the normalized data.
The data set was found in the UCI machine learning database.
This program includes a makefile (g++ for compiling), and can be compiled with the command make. To run, use:
./main data/wine.csvYou can also change the level of verbose output using the "-v" flag. Levels are currently set to use -v, -vv, and -vv.
Remove object files and executables with make clean.
Original Owners:
Forina, M. et al, PARVUS - An Extendible Package for Data Exploration, Classification and Correlation. Institute of Pharmaceutical and Food Analysis and Technologies, Via Brigata Salerno, 16147 Genoa, Italy.
Donor: Stefan Aeberhard, email: stefan '@' coral.cs.jcu.edu.au