public static void Main(string[] args)
{
if (args.Length == 3)
{
ROOT_DIR = args[0];
NUM_TIMESERIES = int.Parse(args[1]);
NEWSPLITPOLICY = bool.Parse(args[2]);
}
InitalizeGlobalSettings();
//BaseIndex();
// Test 1 - Reading from GeneratedRawDataIndex
/*Index<RawDataFormat> testIndex = Index<RawDataFormat>.Load(Globals.IndexRootDir);
* Double[] query = Util.RandomWalk(2);
* IndexEntry result = testIndex.ApproximateSearch(query);
* List<RawDataFormat> timeSeries = testIndex.ReturnDataFormatFromTermEntry((TermEntry)result);*/
// Test 2 - Reading from FileRawDataLoader
//const string DATA_FILE = @"/Users/ppo/Documents/Thesis/test";
const string DATA_FILE = @"/Users/ppo/Dropbox/Thesis/datasets/SonyAIBORobotSurfaceII/SonyAIBORobotSurfaceII_TRAIN_SHAPELET_32_space";
Index <RawShapeletFormat> anotherIndex = new Index <RawShapeletFormat>(0, new IndexOptions("root"));
DataLoader dl = new FromFileRawShapeletLoader(anotherIndex, DATA_FILE, 10000);
InsertTimeSeries(dl);
Index <RawShapeletFormat> .Save(Globals.IndexRootDir, anotherIndex);
Index <RawShapeletFormat> loadBack = Index <RawShapeletFormat> .Load(Globals.IndexRootDir);
OutputIndex(loadBack, "root");
using (StreamWriter sw = new StreamWriter(Path.Combine(Globals.IndexRootDir, "indexOutput.csv"))) {
sw.Write(output);
}
/*foreach (IndexEntry i in anotherIndex.GetIndexEntries()) {
* if (i is TermEntry) {
* ((TermEntry)i).ToString();
* } else {
* foreach (IndexEntry j in ((SplitEntry<RawShapeletFormat>)i).GetIndexEntries())
* {
* if (j is TermEntry)
* {
* ((TermEntry)j).ToString();
* } else {
* ((SplitEntry<RawShapeletFormat>)j).GetIndexEntries();
* }
* }
* }
* }*/
for (int i = 0; i < 100; i++)
{
Double[] query = Util.RandomWalk(4);
IndexEntry result = anotherIndex.ApproximateSearch(query);
List <RawShapeletFormat> timeSeries = anotherIndex.ReturnDataFormatFromTermEntry((TermEntry)result);
}
//Console.WriteLine("Press Enter to exit program.");
//Console.ReadLine();
}
/// <summary>
/// Builds an index from the TinyImages dataset
/// </summary>
public static void TinyImagesIndex()
{
DateTime startTime = DateTime.Now;
const string DATA_DIR = @"D:\jin\TinyImagesBinary";
Index <Meta1DataFormat> si = new Index <Meta1DataFormat>(0, new IndexOptions("root"));
DataLoader dl = new TinyImagesDataLoader(si, DATA_DIR);
InsertTimeSeries(dl);
Index <Meta1DataFormat> .Save(Globals.IndexRootDir, si);
Index <Meta1DataFormat> si2 = Index <Meta1DataFormat> .Load(Globals.IndexRootDir);
DateTime endConstructionTime = DateTime.Now;
Console.WriteLine("Index Construction Time: {0}", endConstructionTime - startTime);
}
/// <summary>
/// Builds an index from the DNA dataset
/// </summary>
public static void DnaIndex()
{
DateTime startTime = DateTime.Now;
Util.NormalizationHandler = new Util.Normalize(Util.MeanZero_Normalization);
const string DATA_FILE = @"M:\Datasets\DNA\Dna2Ts\isax2.0experiment\16.mat.dat";
Index <Meta2DataFormat> si = new Index <Meta2DataFormat>(0, new IndexOptions("root"));
DataLoader dl = new DnaDataLoader(si, DATA_FILE);
InsertTimeSeries(dl);
Index <Meta2DataFormat> .Save(Globals.IndexRootDir, si);
Index <Meta2DataFormat> si2 = Index <Meta2DataFormat> .Load(Globals.IndexRootDir);
DateTime endConstructionTime = DateTime.Now;
Console.WriteLine("Index Construction Time: {0}", endConstructionTime - startTime);
}
/// <summary>
/// Builds an index with randomly generated time series
/// </summary>
public static void BaseIndex()
{
DateTime startTime = DateTime.Now;
// index construction
Index <RawDataFormat> si = new Index <RawDataFormat>(0, new IndexOptions("root"));
DataLoader dl = new GeneratedRawDataLoader(si, Globals.TimeSeriesLength, NUM_TIMESERIES, SEED);
InsertTimeSeries(dl);
Console.WriteLine();
Console.WriteLine("Sequential Disk Accesses: " + DiskCost.seqcost);
Console.WriteLine("Random Disk Accesses: " + DiskCost.rancost);
Console.WriteLine("Read Disk Accesses: " + DiskCost.readcost);
Console.WriteLine("Saved cost in buffer: " + DiskCost.savedcost);
Console.WriteLine();
Index <RawDataFormat> .Save(Globals.IndexRootDir, si);
Index <RawDataFormat> si2 = Index <RawDataFormat> .Load(Globals.IndexRootDir);
DateTime endConstructionTime = DateTime.Now;
Console.WriteLine("Index Construction Time: {0}", endConstructionTime - startTime);
// generate some test queries
const int NUM_QUERIES = 10;
List <double[]> queries = new List <double[]>(NUM_QUERIES);
for (int i = 0; i < NUM_QUERIES; i++)
{
queries.Add(Util.RandomWalk(Globals.TimeSeriesLength));
}
// full sequential scan
Console.WriteLine("Performing full sequential scan.");
Console.WriteLine("--------------------------------");
List <IndexFileDist[]> nnInfo = si.KNearestNeighborSequentialScan(10, queries);
Console.WriteLine();
// query results
Console.WriteLine("Performing exact and approximate search.");
Console.WriteLine("----------------------------------------");
int counter = 0;
for (int i = 0; i < NUM_QUERIES; i++)
{
IndexFileDist exactResult;
si.ExactSearch(queries[i], out exactResult);
IndexFileDist approxResult = Index <RawDataFormat> .MinFileEucDist(queries[i],
si.ApproximateSearch(queries[i]).FileName);
Assert.IsTrue(exactResult == nnInfo[i][0]);
if (approxResult == exactResult)
{
counter++;
Console.WriteLine(approxResult);
}
}
Console.WriteLine("{0} approximate results == exact results.", counter);
Console.WriteLine();
}
public static void SearchQualityExperiment()
{
DateTime startTime = DateTime.Now;
// index construction
Index <RawDataFormat> si = new Index <RawDataFormat>(0, new IndexOptions("root"));
DataLoader dl = new GeneratedRawDataLoader(si, Globals.TimeSeriesLength, NUM_TIMESERIES, SEED);
InsertTimeSeries(dl);
Console.WriteLine();
Console.WriteLine("Sequential Disk Accesses: " + DiskCost.seqcost);
Console.WriteLine("Random Disk Accesses: " + DiskCost.rancost);
Console.WriteLine("Read Disk Accesses: " + DiskCost.readcost);
Console.WriteLine("Saved cost in buffer: " + DiskCost.savedcost);
Console.WriteLine();
Index <RawDataFormat> .Save(Globals.IndexRootDir, si);
Index <RawDataFormat> si2 = Index <RawDataFormat> .Load(Globals.IndexRootDir);
DateTime endConstructionTime = DateTime.Now;
Console.WriteLine("Index Construction Time: {0}", endConstructionTime - startTime);
// avg over queries
const int NUM_QUERIES = 100;
List <double[]> queries = new List <double[]>(NUM_QUERIES);
for (int i = 0; i < NUM_QUERIES; i++)
{
queries.Add(Util.RandomWalk(Globals.TimeSeriesLength));
}
// measured metrics
double approxSearchDist = 0;
double approxSearchNodeDist = 0;
double approxSearchNodeSize = 0;
CostCounter exactSearchCosts = new CostCounter();
for (int i = 0; i < queries.Count; ++i)
{
// exact search
IndexFileDist eRes;
exactSearchCosts += si.ExactSearch(queries[i], out eRes);
// approximate search
TermEntry approxNode = si.ApproximateSearch(queries[i]);
double mDist = double.MaxValue;
List <RawDataFormat> nodeEntries = si.ReturnDataFormatFromTermEntry(approxNode);
double sumDists = 0;
foreach (RawDataFormat rd in nodeEntries)
{
double dist = Util.EuclideanDistance(queries[i], rd.GetTimeSeries());
sumDists += dist;
if (dist < mDist)
{
mDist = dist;
}
}
approxSearchDist += mDist;
approxSearchNodeDist += sumDists / nodeEntries.Count;
approxSearchNodeSize += nodeEntries.Count;
}
approxSearchDist /= queries.Count;
approxSearchNodeDist /= queries.Count;
approxSearchNodeSize /= queries.Count;
using (StreamWriter sw = new StreamWriter(Path.Combine(ROOT_DIR, "searchQuality.txt")))
{
string baseFormat = string.Format("{0}:NumTs_{1}:Th_{2}:Wl_{3}:NewPolicy", NUM_TIMESERIES, Globals.IndexNumMaxEntries, Globals.TimeSeriesLength, Globals.NewSplitPolicy);
sw.WriteLine(baseFormat);
sw.WriteLine("ExactSearchNumIO {0}", exactSearchCosts.IO / (double)queries.Count);
sw.WriteLine("ExactSearchNumCalcuations {0}", exactSearchCosts.distance / (double)queries.Count);
sw.WriteLine("ApproxSearchDistance {0}", approxSearchDist);
sw.WriteLine("ApproxSearchAverageNodeDistance {0}", approxSearchNodeDist);
sw.WriteLine("ApproxSearchAverageNodeSize {0}", approxSearchNodeSize);
sw.WriteLine("ValidationString ");
foreach (double[] query in queries)
{
sw.Write("{0} ", query[1]);
}
sw.WriteLine();
}
}
