public void TestFrontendCall()
{
var config = new ConfigParams(tokenId: "10");
var er = new EnrichmentWrapper(config);
var instance = RandomInstance.RandomizeCoordinatesAndSave(20, config, false);
er.SpatialmHGWrapper(instance.Item1.Zip(instance.Item2, (a, b) => new Tuple <double, double, bool>(a.GetDimension(0), a.GetDimension(1), b)).ToList());
}
public void TestFrontendData()
{
var config = new ConfigParams(tokenId: "");
var er = new EnrichmentWrapper(config);
//var instance = RandomInstance.RandomizeCoordinatesAndSave(20, config, false);
var data = File.ReadAllLines(@"zika.csv").Skip(1).Select(l => l.Split(','))
.Select(v => new Tuple <double, double, bool>(double.Parse(v[2]), double.Parse(v[3]), v[1] == "1"))
.ToList();
var res = er.SpatialmHGWrapper(data);
foreach (var r in res.Cast <SpatialmHGResult>())
{
Console.WriteLine(r.pvalue);
}
}
private static List <ISpatialmHGResult> getSpatialmHGResults(List <Tuple <double, double, bool> > points, Dictionary <string, string> parameters)
{
EnrichmentWrapper wrapper = new EnrichmentWrapper(parameters);
return(wrapper.SpatialmHGWrapper(points));
}
public static void CompareExhaustiveWithPivots(int numcoords = 50, int numiter = 500)
{
Config = new ConfigParams("");
#region init
StaticConfigParams.rnd = (Config.ActionList & Actions.Program_RandomConstSeed) != 0 ? new SafeRandom(1) : new SafeRandom();
Config.timer.Start();
#endregion
//Load coordinates and labels
var identities = new List <string>();
var resultPairedDiff = new List <double>();
int victories = 0, ties = 0;
using (var fileout = new StreamWriter(@"pivot_vs_exhaustive.csv"))
for (var instanceIter = 0; instanceIter < numiter; instanceIter++)
{
var coordinates = new List <ICoordinate>();
var labels = new List <bool>();
StaticConfigParams.filenamesuffix = instanceIter.ToString();
Console.WriteLine("File {0}", instanceIter);
var res = Program.RandomizeCoordinatesAndSave(numcoords, false);
coordinates = res.Item1;
labels = res.Item2;
var zeros = labels.Count(l => l == false);
var filterCount = (int)(Config.FilterKFurthestZeros * zeros);
if (filterCount > 0)
{
Console.WriteLine("Filtering {0} far away points", filterCount);
var positives = new List <ICoordinate>();
var negatives = new List <ICoordinate>();
var negIds = new List <int>();
for (var i = 0; i < coordinates.Count; i++)
{
if (labels[i])
{
positives.Add(coordinates[i]);
}
else
{
negatives.Add(coordinates[i]);
negIds.Add(i);
}
}
var negMinDist = new HashSet <int>(negatives.Zip(negIds, (a, b) => new { PosMinDist = positives.Select(p => p.EuclideanDistance(a)).Min(), Id = b })
.OrderByDescending(n => n.PosMinDist).Select(t => t.Id).Take(filterCount));
coordinates = coordinates.Where((a, b) => !negMinDist.Contains(b)).ToList();
labels = labels.Where((a, b) => !negMinDist.Contains(b)).ToList();
numcoords -= filterCount;
}
//Actual work starts here
var ones = labels.Count(l => l);
var linecount = ones * (numcoords - ones);
Config.Cellcount = ((long)linecount * (linecount - 1)) / 2.0 + linecount + 1;
mHGJumper.Initialize(ones, numcoords - ones);
mHGJumper.optHGT = Config.SIGNIFICANCE_THRESHOLD; // / Cellcount; //for bonferonni
//alpha is the Bonferonni (union-bound) corrected significance level
Tesselation T = null;
var ew = new EnrichmentWrapper(Config);
Console.WriteLine(@"Starting work on {0} coordinates with {1} 1's (|cells|={2:n0}, alpha={3}).", numcoords, ones, Config.Cellcount, mHGJumper.optHGT);
var instanceData = coordinates.Zip(labels, (a, b) => new Tuple <double, double, bool>(a.GetDimension(0), a.GetDimension(1), b)).ToList();
var resultsExhaustive = ew.SpatialmHGWrapper(instanceData).Select(v => (SpatialmHGResult)v).First();
var resultsPivot = ew.mHGPivotWrapper(instanceData).Select(v => (SpatialmHGResult)v).First();
fileout.WriteLine($"{resultsExhaustive.pvalue}, {resultsPivot.pvalue}");
if (resultsExhaustive.pvalue < resultsPivot.pvalue)
{
victories++;
}
else if (resultsExhaustive.pvalue == resultsPivot.pvalue)
{
ties++;
}
else
{
Console.WriteLine($"Debug me");
}
resultPairedDiff.Add(Math.Log10(resultsPivot.pvalue) - Math.Log10(resultsExhaustive.pvalue));
}
Console.WriteLine($"Out of {numiter} iterations, spatial enrichment won in {victories} and tied in {ties}.");
Console.WriteLine("Total elapsed time: {0:g}.\nPress any key to continue.", Config.timer.Elapsed);
File.WriteAllLines("experiment_pvaldiffs.txt", resultPairedDiff.Select(v => v.ToString()).ToArray());
Console.ReadKey();
}
/// <summary>
/// 3d subsample from 50 points 20. run 100 times. compare to opt.
/// </summary>
/// <param name="numcoords"></param>a
/// <param name="numiter"></param>
public static List <double> CompareExahustiveWithSubsamplingInput(int numcoords = 50, int subsampleSize = 20, int numiter = 100, string suffix = "0")
{
Config = new ConfigParams("");
#region init
StaticConfigParams.rnd = (Config.ActionList & Actions.Program_RandomConstSeed) != 0 ? new SafeRandom(1) : new SafeRandom();
Config.timer.Start();
#endregion
//Load coordinates and labels
var resultPairedDiff = new List <double>();
var extraAnalyses = new List <string>();
int victories = 0, ties = 0;
var coordinates = new List <ICoordinate>();
var labels = new List <bool>();
Program.Config = Config;
StaticConfigParams.filenamesuffix = suffix;
var res = Program.RandomizeCoordinatesAndSave(numcoords, true);
coordinates = res.Item1;
labels = res.Item2;
var zeros = labels.Count(l => l == false);
var filterCount = (int)(Config.FilterKFurthestZeros * zeros);
if (filterCount > 0)
{
Console.WriteLine("Filtering {0} far away points", filterCount);
var positives = new List <ICoordinate>();
var negatives = new List <ICoordinate>();
var negIds = new List <int>();
for (var i = 0; i < coordinates.Count; i++)
{
if (labels[i])
{
positives.Add(coordinates[i]);
}
else
{
negatives.Add(coordinates[i]);
negIds.Add(i);
}
}
var negMinDist = new HashSet <int>(negatives.Zip(negIds, (a, b) => new { PosMinDist = positives.Select(p => p.EuclideanDistance(a)).Min(), Id = b })
.OrderByDescending(n => n.PosMinDist).Select(t => t.Id).Take(filterCount));
coordinates = coordinates.Where((a, b) => !negMinDist.Contains(b)).ToList();
labels = labels.Where((a, b) => !negMinDist.Contains(b)).ToList();
numcoords -= filterCount;
}
var instanceDataCoords = coordinates.Zip(labels, (a, b) => new Tuple <ICoordinate, bool>(a, b)).ToList();
var instanceData = coordinates.Zip(labels, (a, b) => new Tuple <double, double, bool>(a.GetDimension(0), a.GetDimension(1), b)).ToList();
Config.SKIP_SLACK = -1000;
var ew = new EnrichmentWrapper(Config);
var resultsExhaustive = ew.SpatialmHGWrapper(instanceData).Select(v => (SpatialmHGResult)v).First();
//Actual work starts here
var ones = labels.Count(l => l);
var linecount = ones * (numcoords - ones);
Config.Cellcount = ((long)linecount * (linecount - 1)) / 2.0 + linecount + 1;
mHGJumper.Initialize(ones, numcoords - ones);
mHGJumper.optHGT = Config.SIGNIFICANCE_THRESHOLD; // / Cellcount; //for bonferonni
//alpha is the Bonferonni (union-bound) corrected significance level
using (var fileout = new StreamWriter($"sample_vs_exhaustive_{suffix}.csv"))
for (var instanceIter = 1; instanceIter < numiter; instanceIter++)
{
StaticConfigParams.filenamesuffix = instanceIter.ToString();
Console.WriteLine("File {0}", instanceIter);
var sampleCoords = coordinates
.Zip(labels, (a, b) => new { Coords = a, Labels = b, Rand = StaticConfigParams.rnd.Next() })
.OrderBy(v => v.Rand).Take(subsampleSize).ToList();
while (sampleCoords.All(v => v.Labels) || !sampleCoords.Any(v => v.Labels))
{
sampleCoords = coordinates
.Zip(labels, (a, b) => new { Coords = a, Labels = b, Rand = StaticConfigParams.rnd.Next() })
.OrderBy(v => v.Rand).Take(subsampleSize).ToList();
}
if (StaticConfigParams.WriteToCSV)
{
Generics.SaveToCSV(sampleCoords.Select(t => t.Coords.ToString() + "," + Convert.ToDouble(t.Labels)),
$@"coords_{StaticConfigParams.filenamesuffix}.csv");
}
ones = sampleCoords.Count(l => l.Labels);
linecount = ones * (subsampleSize - ones);
Config.Cellcount = ((long)linecount * (linecount - 1)) / 2.0 + linecount + 1;
Console.WriteLine(@"Starting work on {0} coordinates with {1} 1's (|cells|={2:n0}, alpha={3}).", numcoords, ones, Config.Cellcount, mHGJumper.optHGT);
mHGJumper.optHGT = Config.SIGNIFICANCE_THRESHOLD;
Tesselation T = new Tesselation(sampleCoords.Select(v => (Coordinate)v.Coords).ToList(), sampleCoords.Select(v => v.Labels).ToList(), null, Config)
{
ProjectedFrom = coordinates,
SourceLabels = labels.ToArray()
};
var topResults = T.GradientSkippingSweep(numStartCoords: 20, numThreads: Environment.ProcessorCount - 1).First();
Line.Reset();
if (resultsExhaustive.pvalue < topResults.mHG.Item1)
{
victories++;
}
else if (resultsExhaustive.pvalue == topResults.mHG.Item1)
{
ties++;
}
else
{
Console.WriteLine($"Debug me");
}
var pdiff = Math.Log10(topResults.mHG.Item1) - Math.Log10(resultsExhaustive.pvalue);
resultPairedDiff.Add(pdiff);
mHGJumper.optHGT = Config.SIGNIFICANCE_THRESHOLD;
Console.Write($"Uniform grid strategy @{Config.Cellcount} pivots... ");
var uniformGridFactory = new Gridding();
uniformGridFactory.GeneratePivotGrid(Convert.ToInt64(Config.Cellcount));
var uniformGridPivotlst = uniformGridFactory.GetPivots().ToList();
var uniformGridPivot = uniformGridPivotlst.AsParallel().Max(p => - Math.Log10(EnrichmentAtPivot(instanceDataCoords, p)));
Console.WriteLine($"p={uniformGridPivot:e}");
Console.Write($"Empirical grid strategy @{Config.Cellcount} pivots... ");
mHGJumper.optHGT = Config.SIGNIFICANCE_THRESHOLD;
var empiricalGridFactory = new Gridding();
empiricalGridFactory.GenerateEmpricialDensityGrid(Convert.ToInt64(Config.Cellcount), instanceDataCoords);
var empiricalGridPivotlst = empiricalGridFactory.GetPivots().ToList();
var empiricalGridPivot = empiricalGridPivotlst.AsParallel().Max(p => - Math.Log10(EnrichmentAtPivot(instanceDataCoords, p)));
Console.WriteLine($"p={empiricalGridPivot:e}");
//extraAnalyses.Add($"{-Math.Log10(resultsExhaustive.pvalue)}, {-Math.Log10(topResults.mHG.Item1)}, {uniformGridPivot}, {empiricalGridPivot}");
fileout.WriteLine($"{-Math.Log10(resultsExhaustive.pvalue)}, {-Math.Log10(topResults.mHG.Item1)}, {uniformGridPivot}, {empiricalGridPivot}");
}
Console.WriteLine($"Out of {numiter} iterations, spatial enrichment won in {victories} and tied in {ties}.");
Console.WriteLine("Total elapsed time: {0:g}.\nPress any key to continue.", Config.timer.Elapsed);
//File.WriteAllLines($"experiment_pvaldiffs_{suffix}.txt", resultPairedDiff.Select(v => v.ToString()).ToArray());
//File.WriteAllLines($"experimentsAll_{suffix}.txt", extraAnalyses);
return(resultPairedDiff);
}
static void Main(string[] args)
{
var options = new CommandlineParameters();
var isValid = Parser.Default.ParseArgumentsStrict(args, options);
//args = new[] {@"c:\Users\shaybe\Dropbox\Thesis-PHd\SpatialEnrichment\Datasets\usStatesBordersData.csv"};
//args = new[] { @"c:\Users\shaybe\Dropbox\Thesis-PHd\SpatialEnrichment\Caulobacter\transferases\acetyltransferase.csv" };
var numcoords = 300;
Config = new ConfigParams("");
if ((Config.ActionList & Actions.Experiment_ComparePivots) != 0)
{
Console.WriteLine(@"Running pivot comparison experiment");
Experiments.CompareExhaustiveWithPivots(numcoords, numiter: 30);
return;
}
if ((Config.ActionList & Actions.Experiment_SampleLines) != 0)
{
Console.WriteLine(@"Running sampling comparison experiment");
var subsamples = new[] { 10, 20, 30 };
var population = new[] { 40, 60, 100 };
var counter = 0;
foreach (var nu in subsamples)
{
foreach (var N in population)
{
Experiments.CompareExahustiveWithSubsamplingInput(N, nu, 50, counter++.ToString());
}
}
return;
}
if (Config.SKIP_SLACK != 0)
{
Console.WriteLine(@"Warning! Current configuration uses CONST_SKIP_SLACK={0}", Config.SKIP_SLACK);
}
if (StaticConfigParams.WriteToCSV)
{
Console.WriteLine(@"Warning! Current configuration writes cells to CSV - this is SLOW.");
}
#region init
StaticConfigParams.rnd = (Config.ActionList & Actions.Program_RandomConstSeed) != 0 ? new SafeRandom(1) : new SafeRandom();
Config.timer.Start();
#endregion
foreach (var dir in new List <string>()
{
"Cells", "Planes"
})
{
var di = new DirectoryInfo(dir);
if (!di.Exists)
{
di.Create();
}
foreach (FileInfo file in di.GetFiles())
{
file.Delete();
}
}
foreach (var filemask in new List <string>()
{
"lines_*.csv", "coordSample_*.csv "
})
{
FileInfo[] taskFiles = new DirectoryInfo(Directory.GetCurrentDirectory()).GetFiles(filemask);
foreach (FileInfo file in taskFiles)
{
file.Delete();
}
}
//Load coordinates and labels
var infile = Path.GetFileNameWithoutExtension(args.Length > 0?args[0]:"");
var identities = new List <string>();
for (var instanceIter = 0; instanceIter < 1; instanceIter++)
{
var coordinates = new List <ICoordinate>();
List <bool> labels = null;
StaticConfigParams.filenamesuffix = instanceIter.ToString();
Console.WriteLine("File {0}", instanceIter);
if (args.Length > 0)
{
if (File.Exists(args[0]))
{
var res = LoadCoordinatesFromFile(args, ref numcoords, identities);
coordinates = res.Item1;
labels = res.Item2;
}
else
{
throw new ArgumentException("Input file not found!");
}
}
else
{
var res = RandomizeCoordinatesAndSave(numcoords);
coordinates = res.Item1;
labels = res.Item2;
}
var zeros = labels.Count(l => l == false);
var filterCount = (int)(Config.FilterKFurthestZeros * zeros);
if (filterCount > 0)
{
Console.WriteLine("Filtering {0} far away points", filterCount);
var positives = new List <ICoordinate>();
var negatives = new List <ICoordinate>();
var negIds = new List <int>();
for (var i = 0; i < coordinates.Count; i++)
{
if (labels[i])
{
positives.Add(coordinates[i]);
}
else
{
negatives.Add(coordinates[i]);
negIds.Add(i);
}
}
var negMinDist = new HashSet <int>(negatives.Zip(negIds, (a, b) => new { PosMinDist = positives.Select(p => p.EuclideanDistance(a)).Min(), Id = b })
.OrderByDescending(n => n.PosMinDist).Select(t => t.Id).Take(filterCount));
coordinates = coordinates.Where((a, b) => !negMinDist.Contains(b)).ToList();
labels = labels.Where((a, b) => !negMinDist.Contains(b)).ToList();
numcoords -= filterCount;
}
//Actual work starts here
var ones = labels.Count(l => l);
var linecount = ones * (numcoords - ones);
Config.Cellcount = ((long)linecount * (linecount - 1)) / 2.0 + linecount + 1;
//Look at lazy caretaker numbers. Note, we don't actually cover open cells
//so its -linecount as each line has two open cells on either side of it,
//and each open cell is made up of two lines.
mHGJumper.Initialize(ones, numcoords - ones);
mHGJumper.optHGT = Config.SIGNIFICANCE_THRESHOLD;// / Cellcount; //for bonferonni
//alpha is the Bonferonni (union-bound) corrected significance level
//Debugging.debug_mHG(numcoords,ones);
Tesselation T = null;
var coordType = coordinates.First().GetType();
var ew = new EnrichmentWrapper(Config);
List <ISpatialmHGResult> results = null;
if (coordType == typeof(Coordinate3D))
{
Config.Cellcount += MathExtensions.Binomial(linecount, 3);
Console.WriteLine(@"Projecting 3D problem to collection of 2D {0} coordinates with {1} 1's (|cells|={2:n0}, alpha={3}).", numcoords, ones, Config.Cellcount, mHGJumper.optHGT);
results = ew.SpatialmHGWrapper3D(coordinates.Zip(labels,
(a, b) => new Tuple <double, double, double, bool>(a.GetDimension(0), a.GetDimension(1),
a.GetDimension(2), b)).ToList(), options.BatchMode);
}
else if (coordType == typeof(Coordinate))
{
Console.WriteLine(@"Starting work on {0} coordinates with {1} 1's (|cells|={2:n0}, alpha={3}).", numcoords, ones, Config.Cellcount, mHGJumper.optHGT);
results = ew.SpatialmHGWrapper(coordinates.Zip(labels, (a, b) =>
new Tuple <double, double, bool>(a.GetDimension(0), a.GetDimension(1), b)).ToList());
}
for (var resid = 0; resid < results.Count; resid++)
{
results[resid].SaveToCSV($@"Cells\{infile}_Cell_{resid}_{StaticConfigParams.filenamesuffix}.csv");
}
using (var outfile = new StreamWriter($"{infile}_mhglist_{StaticConfigParams.filenamesuffix}.csv"))
foreach (var res in Config.mHGlist.Where(t => t != null))
{
outfile.WriteLine("{0},{1}", res.Item2, res.Item1);
}
if (options.BatchMode)
{
AzureBatchExecution.UploadFileToContainer($"{infile}_mhglist_{StaticConfigParams.filenamesuffix}.csv", options.SaasUrl);
for (var resid = 0; resid < results.Count; resid++)
{
AzureBatchExecution.UploadFileToContainer($@"Cells\{infile}_Cell_{resid}_{StaticConfigParams.filenamesuffix}.csv", options.SaasUrl);
}
}
}
//Finalize
if (args.Length == 0 || Debugger.IsAttached)
{
Console.WriteLine("Total elapsed time: {0:g}.\nPress any key to continue.", Config.timer.Elapsed);
Console.ReadKey();
}
}