//[TestMethod]
public void KNN_Does_Its_Stuff()
{
const int TRAIN_LIMIT = 50;
const int TEST_LIMIT = 500;
var parser = new CIFAR10Parser();
IDataFormat dataFormat = parser.Format;
//gather the data
IEnumerable <object[]> dataInput = parser.ExtractFeaturesAndLabels(File.Open(Path.Combine(basePath, "dataBatchFull.bin"), FileMode.Open), TRAIN_LIMIT);
//clear and repopulate the datastore
IDataStore dataStore = new DataStoreFactory().CreateOrReplace("TestJack", dataFormat);
dataStore.Clear();
var addTask = dataStore.AddLabelledRow(dataInput);
Task.WaitAll(addTask);
IPredictiveModelFactory modelFactory = new PredictiveModelFactory();
Dictionary <string, string> parameters = new Dictionary <string, string>();
parameters.Add("k", "5");
parameters.Add("numberOfClasses", "10");
IPredictiveModel model = modelFactory.Create("KNearestNeighbour", parameters);
object[][] data = dataStore.GetLabelled().Result.Take(TRAIN_LIMIT).ToArray();
var features = data.GetFeatures <double>(dataFormat);
//we know our ids are sequential from 1 so generate a list of ids for ourselves
var ids = Enumerable.Range(1, TRAIN_LIMIT).Select(i => (long)i);
var labels = dataStore.GetLabelById(ids).Result.Take(TRAIN_LIMIT).Select(l => Convert.ToDouble(l));
model.Train(features, labels);
var testData = parser.ExtractFeatureValues(File.Open(Path.Combine(basePath, "dataBatch5.bin"), FileMode.Open)).Take(TEST_LIMIT);
var testLabels = parser.ExtractLabels(File.Open(Path.Combine(basePath, "dataBatch5_2.bin"), FileMode.Open)).Take(TEST_LIMIT).ToDictionary(k => k.Key, v => v.Value);
int correct = 0;
int incorrect = 0;
object[][] testArr = testData.ToArray();
foreach (object[] row in testArr)
{
var prediction = model.Compute(dataFormat.GetFeatures <double>(row).ToArray());
if (System.Convert.ToInt32(prediction) == System.Convert.ToInt32(testLabels[System.Convert.ToInt32(row[0])]))
{
correct++;
}
else
{
incorrect++;
}
}
Assert.IsTrue(correct > 0);
}
