private static void Main(string[] args)
{
var app = new CommandLineApplication(false);
app.Name = nameof(MNIST);
app.Description = "The exsample program for MNIST";
app.HelpOption("-h|--help");
var quietArgument = app.Argument("quiet", "Suppress output of LIBSVM");
var outputOption = app.Option("-o|--output", "output path for trained model", CommandOptionType.SingleValue);
app.OnExecute(() =>
{
if (quietArgument.Value != null)
{
LibSvm.SetPrintFunction(null);
}
var output = outputOption.Value();
// Load training data and test data set
using (var train = Problem.FromFile("mnist"))
using (var test = Problem.FromFile("mnist.t"))
{
// Configure parameter
var param = new Parameter
{
SvmType = SvmType.CSVC,
KernelType = KernelType.Linear,
Gamma = 0.05d,
C = 5d,
CacheSize = 1000,
Degree = 3,
Coef0 = 0,
Nu = 0.5,
Epsilon = 1e-3,
P = 0.1,
Shrinking = true,
Probability = false,
WeightLabel = new int[0],
Weight = new double[0]
};
var message = LibSvm.CheckParameter(train, param);
if (!string.IsNullOrWhiteSpace(message))
{
Console.WriteLine($"Error: {message} for train problem");
return(-1);
}
// Train training data
var sw = new Stopwatch();
sw.Start();
using (var model = LibSvm.Train(train, param))
{
sw.Stop();
if (!string.IsNullOrWhiteSpace(output))
{
Model.Save(output, model);
}
var correct = 0;
var total = 0;
var x = test.X;
var y = test.Y;
for (var i = 0; i < test.Length; i++)
{
// Get vector from test data
var array = x[i];
// Get classification result (returns label)
var ret1 = (int)LibSvm.Predict(model, array);
if (ret1 == (int)y[i])
{
correct++;
}
total++;
}
Console.WriteLine($"Accuracy: {correct / (double)total * 100}%, Elapsed: {sw.ElapsedMilliseconds / 1000}s");
}
}
return(0);
});
app.Execute(args);
}
private static void Main()
{
// Create random data
var r = new Random();
const int trainCount = 500;
const int testCount = 100;
var temp = Path.GetTempPath();
var trainDic = new StringBuilder();
var testDic = new StringBuilder();
var testDicAns = new List <int>();
for (var l = 0; l < 10; l++)
{
for (var i = 0; i < trainCount; i++)
{
// Create decimal value 0 or greater but less than 10
var v = r.NextDouble() + l;
trainDic.AppendLine($"{l} 1:{v}");
}
for (var i = 0; i < testCount; i++)
{
// Create decimal value 0 or greater but less than 10
var v = r.NextDouble() + l;
testDic.AppendLine($"{l} 1:{v}");
testDicAns.Add(l);
}
}
var tempTrainPath = Path.Combine(temp, "train");
var tempTestPath = Path.Combine(temp, "test");
using (var fs = new FileStream(tempTrainPath, FileMode.Create, FileAccess.Write, FileShare.Write))
using (var sw = new StreamWriter(fs, Encoding.ASCII))
sw.Write(trainDic.ToString());
using (var fs = new FileStream(tempTestPath, FileMode.Create, FileAccess.Write, FileShare.Write))
using (var sw = new StreamWriter(fs, Encoding.ASCII))
sw.Write(testDic.ToString());
// Load training data and test data set
using (var train = Problem.FromFile(tempTrainPath))
using (var test = Problem.FromFile(tempTestPath))
{
// Configure parameter
var param = new Parameter
{
SvmType = SvmType.CSVC,
KernelType = KernelType.RBF,
Gamma = 0.05d,
C = 5d,
CacheSize = 100,
Degree = 3,
Coef0 = 0,
Nu = 0.5,
Epsilon = 1e-3,
P = 0.1,
Shrinking = true,
Probability = false,
WeightLabel = new int[0],
Weight = new double[0]
};
var message = LibSvm.CheckParameter(train, param);
if (!string.IsNullOrWhiteSpace(message))
{
Console.WriteLine($"Error: {message} for train problem");
return;
}
message = LibSvm.CheckParameter(test, param);
if (!string.IsNullOrWhiteSpace(message))
{
Console.WriteLine($"Error: {message} for test problem");
return;
}
// Train training data
using (var model = LibSvm.Train(train, param))
{
var correct = 0;
var total = 0;
var x = test.X;
for (var i = 0; i < test.Length; i++)
{
// Get vector from test data
var array = x[i];
// Get classification result (returns label)
var ret1 = (int)LibSvm.Predict(model, array);
if (ret1 == testDicAns[i])
{
correct++;
}
total++;
}
Console.WriteLine($"Accuracy: {correct / (double)total * 100}%");
}
}
}
private static void Main()
{
// Create random data
var r = new Random();
const int trainCount = 500;
const int testCount = 100;
var trainNodes = new List <Node[]>();
var trainLabels = new List <double>();
var testNodes = new List <Node[]>();
var testLabels = new List <double>();
for (var l = 0; l < 2; l++)
{
for (var i = 0; i < trainCount; i++)
{
// Create decimal value 0 or greater but less than 2
var v = r.NextDouble() + l;
trainNodes.Add(new[] { new Node {
Index = 0, Value = v
} });
trainLabels.Add(l);
}
for (var i = 0; i < testCount; i++)
{
// Create decimal value 0 or greater but less than 2
var v = r.NextDouble() + l;
testNodes.Add(new[] { new Node {
Index = 0, Value = v
} });
testLabels.Add(l);
}
}
// Load training data and test data set
using (var train = Problem.FromSequence(trainNodes, trainLabels))
using (var test = Problem.FromSequence(testNodes, testLabels))
{
// Configure parameter
var param = new Parameter
{
SvmType = SvmType.CSVC,
KernelType = KernelType.RBF,
Gamma = 0.05d,
C = 5d,
CacheSize = 100,
Degree = 3,
Coef0 = 0,
Nu = 0.5,
Epsilon = 1e-3,
P = 0.1,
Shrinking = true,
Probability = false,
WeightLabel = new int[0],
Weight = new double[0]
};
var message = LibSvm.CheckParameter(train, param);
if (!string.IsNullOrWhiteSpace(message))
{
Console.WriteLine($"Error: {message} for train problem");
return;
}
message = LibSvm.CheckParameter(test, param);
if (!string.IsNullOrWhiteSpace(message))
{
Console.WriteLine($"Error: {message} for test problem");
return;
}
// Train training data
using (var model = LibSvm.Train(train, param))
{
var correct = 0;
var total = 0;
var x = test.X;
for (var i = 0; i < test.Length; i++)
{
// Get vector from test data
var array = x[i];
// Get classification result (returns label)
var ret1 = (int)LibSvm.Predict(model, array);
if (ret1 == (int)testLabels[i])
{
correct++;
}
total++;
}
Console.WriteLine($"Accuracy: {correct / (double)total * 100}%");
}
}
}