public static List <Vector <double> > Test(HopfieldNetwork network, List <Vector <double> > patterns, int testPattern, double noise)
{
Random rand = new Random();
var input = patterns[testPattern].Map((v) => rand.NextDouble() < noise ? v * -1.0 : v);
return(network.GetPattern(input));
}
public static async Task SaveEnergyMetrics(HopfieldNetwork network, List <Vector <double> > patterns, int repetitions, int testPattern, double noise)
{
var energyValues = new List <List <double> >();
Random rand = new Random();
for (var i = 0; i < repetitions; i++)
{
var input = patterns[testPattern].Map((v) => rand.NextDouble() < noise ? v * -1.0 : v);
energyValues.Add(network.GetEnergy(input));
}
for (var i = 0; i < energyValues.Count; i++)
{
await File.WriteAllLinesAsync($"energy_{i}.csv", energyValues[i].Select((v, index) => $"{index},{v}"));
}
}
public static Task SaveNoiseAccuracyMetrics(HopfieldNetwork network, List <Vector <double> > patterns, int repetitions, int testPattern)
{
var accuracy = new List <double>();
Random rand = new Random();
for (float n = 0; n < 1.1; n += 0.1f)
{
var success = 0;
for (int i = 0; i < repetitions; i++)
{
var result = Test(network, patterns, testPattern, n);
if (patterns.IndexOf(result[result.Count - 1]) == testPattern)
{
success++;
}
}
accuracy.Add(success / (double)repetitions);
}
return(File.WriteAllLinesAsync("accuracy.csv", accuracy.Select((v, index) => $"{index * 0.1},{v}")));
}
