Beispiel #1
0
        static void TestXor <U>(bool summary = false, int epochs = 50, int displayEpochs = 25)
        {
            Console.WriteLine("Hello World, MLP on Xor Dataset.");

            var(trainX, trainY) = ImportDataset.XorDataset <U>();
            var net = new Network <U>(new SGD <U>(0.1f), new MeanSquaredLoss <U>(), new RoundAccuracy <U>());

            net.AddLayer(new DenseLayer <U>(8, inputShape: 2));
            net.AddLayer(new TanhLayer <U>());
            net.AddLayer(new DenseLayer <U>(1));
            net.AddLayer(new SigmoidLayer <U>());

            if (summary)
            {
                net.Summary();
            }

            net.Fit(trainX, trainY, epochs, displayEpochs: displayEpochs);

            if (summary)
            {
                var yp = net.Forward(ndX);
                for (int k = 0; k < 4; ++k)
                {
                    Console.WriteLine($"[{X[k].Glue()}] = [{y[k][0]}] -> {yp.Data[k]:0.000000}");
                }
            }

            Console.WriteLine();
        }
Beispiel #2
0
        static void TestDigits <U>(bool summary = false, int epochs = 50, int displayEpochs = 25, int batchsize = 100)
        {
            Console.WriteLine("Hello World, MLP on Digits Dataset.");

            (var trainX, var trainY, var testX, var testY) = ImportDataset.DigitsDataset <U>(ratio: 0.9);
            var net = new Network <U>(new SGD <U>(0.025f, 0.2f), new CrossEntropyLoss <U>(), new ArgmaxAccuracy <U>());

            net.AddLayer(new DenseLayer <U>(32, inputShape: 64));
            net.AddLayer(new SigmoidLayer <U>());
            net.AddLayer(new DenseLayer <U>(10));
            net.AddLayer(new SoftmaxLayer <U>());

            if (summary)
            {
                net.Summary();
            }

            net.Fit(trainX, trainY, testX, testY, epochs: epochs, batchSize: batchsize, displayEpochs: displayEpochs);

            Console.WriteLine();
        }
Beispiel #3
0
        static void TestIris <U>(bool summary = false, int epochs = 50, int displayEpochs = 25, int batchsize = 10)
        {
            Console.WriteLine("Hello World, MLP on Iris Dataset.");

            (var trainX, var trainY, var testX, var testY) = ImportDataset.IrisDataset <U>(ratio: 0.8);
            var net = new Network <U>(new SGD <U>(0.025f, 0.2f), new MeanSquaredLoss <U>(), new ArgmaxAccuracy <U>());

            net.AddLayer(new DenseLayer <U>(5, inputShape: 4));
            net.AddLayer(new TanhLayer <U>());
            net.AddLayer(new DenseLayer <U>(3));
            net.AddLayer(new SigmoidLayer <U>());

            if (summary)
            {
                net.Summary();
            }

            net.Fit(trainX, trainY, epochs, batchSize: batchsize, displayEpochs: displayEpochs);
            net.Test(testX, testY);

            Console.WriteLine();
        }