- 8623910 - Adriano dos Santos Rodrigues da Silva
- 8516883 - Caio Vinicius Marques Teixeira
- 8598631 - Guilherme Hernandes do Nascimento
- 8623865 - João Pedro Nardari dos Santos
#1. How to compile and run the neural networks Open the RedesNeurais folder and run the following commands:
javac -cp ".\libs\commons-cli-1.3.jar;" .\*.java
java -cp "libs\commons-cli-1.3.jar;" MLPDigits + Argumentos
java -cp "libs\commons-cli-1.3.jar;" LVQDigits + Argumentos
Obs: You should use the cp argument to include the commons-cli-1.3.jar library
Obs2: Look below about how to use parameters
#2. MLP ##2.1 MLP parameters usage: MLPDigits [-bias] [-hc ] [-ic ] [-init ] [-lc ] [-load ] [-lr ] [-oc ] [-save ] [-testlog ] [-tn ] [-trainlog ] [-tt ] [-vl ] -bias use bias -hc number of hidden nodes -ic number of input nodes -init neuron initialization methods (RANDOM or ZERO) -lc number of hidden layers -load path for file containing existing LVQ network -lr learning rate -oc number of output nodes -save save the neural network, insert the file path -testlog save a log file, insert the file path -tn training data set path -trainlog save training log file, insert the file path -tt test data set path -vl validation data set path
##2.2 MLP Examples 2.2.1 Create a new MLP network, train (with validation), test, generate and save log file
java -cp "libs\commons-cli-1.3.jar;" MLPDigits
-init RANDOM
-lr 0.5
-bias
-lc 1
-ic 61
-oc 4
-hc 40
-tn "optdigits.norm.cortado.tra"
-tt "optdigits.norm.cortado.tes"
-vl "optdigits.norm.cortado.val"
-trainlog "trainningLogMLPDigits.csv"
-testlog "testLogMLPDigits.csv"
-save "mlpNetwork.mlp"
2.2.2 Load a MLP network, test and save output data
java -cp "libs\commons-cli-1.3.jar;" MLPDigits
-load "mlpNetwork.mlp"
-tt "optdigits.norm.cortado.tes"
-testlog "testLogMLPDigits.csv"
#3. LVQ ##3.1 LVQ Parameters usage: LVQDigits [-distance ] [-init ] [-load ] [-lr ] [-nc ] [-rr ] [-save ] [-testlog ] [-tn ] [-trainlog ] [-tt ] [-vl ] -distance length metric mode (EUCLIDEAN or MANHATTAN) -init neuron initialization method (RANDOM , FIRST_VALUES ou ZERO) -load File path for LVQ network -lr learning rate -nc number of neurons -rr reduction rate -save save the neural network, insert the file path -testlog save a test log file, insert the file path -tn training data set path -trainlog save training log file, insert the file path -tt test data set path -vl validation data set path
##3.2 LVQ Examples 3.2.1 Create a new LVQ network, train (with validation), test, generate and save log file
java -cp "libs\commons-cli-1.3.jar;" LVQDigits
-init FIRST_VALUES
-distance EUCLIDEAN
-lr 0.001
-rr 30
-nc 4
-tn "optdigits.norm.cortado.tra"
-tt "optdigits.norm.cortado.tes"
-vl "optdigits.norm.cortado.val"
-trainlog "trainningLogLVQDigits.csv"
-testlog "testLogLVQDigits.csv"
-save "lvqNetwork.lvq"
3.2.2 Load a LVQ network, test and save output data
java -cp "libs\commons-cli-1.3.jar;" LVQDigits
-load "lvqNetwork.lvq"
-tt "optdigits.norm.cortado.tes"
-testlog "testLogLVQDigits.csv"
=====================================================================================