static void Main(string[] args)
{
var dataset = "user.emmat.mc15_13TeV.361022.Pythia8EvtGen_A14NNPDF23LO_jetjet_JZ2W.merge.AOD.e3668_s2576_s2132_r6765_r6282__EXOT15_v3_EXT0";
var jobFiles = GRIDJobs.FindJobFiles("DiVertAnalysis", 3, dataset);
var events = QueryablerecoTree.CreateQueriable(jobFiles);
events.CleanupQuery = false; // Set to keep C++ around
events.IgnoreQueryCache = true; // Set to ignore query cache for speed
using (var f = new FutureTFile("S2.root"))
{
var count = events.FutureCount();
var jets = events
.SelectMany(e => e.Jet)
.Where(j => Math.Abs(j.eta) < 1.5);
jets
.FuturePlot("pt", "Jet pT; pT [GeV]", 100, 0.0, 100.0, j => j.pT)
.Save(f);
jets
.FuturePlot("eta", "Jet eta; eta", 100, -2.0, 2.0, j => j.eta)
.Save(f);
Console.WriteLine($"Saw {count.Value} events.");
}
}
static void Main(string[] args)
{
WriteLine("Finding the files");
var backgroundEvents = Files.GetJZ(2).Select(e => e.Data);
//var signalHV125pi15Events = Files.Get125pi15();
//var signalHV125pi40Events = Files.Get125pi40();
var signalHV600pi100Events = Files.Get600pi150lt9m();
//var signalGet200pi25lt5mEvents = Files.Get200pi25lt5m();
//
// Do a simple cut training here
//
var t = TrainingIQueriable(signalHV600pi100Events, true)
.AsSignal("HV600pi100")
.Background(TrainingIQueriable(backgroundEvents, false), "J2Z")
.IgnoreVariables(p => p.lowestPtTrack);
var mCuts = t.AddMethod(ROOTNET.Interface.NTMVA.NTypes.EMVA.kCuts, "SimpleCuts")
.Option("!H:V:FitMethod=MC:EffSel:SampleSize=200000:VarTransform=Decorrelate")
.ParameterOption(p => p.logR, "VarProp", "FSmart")
.ParameterOption(p => p.nTracks, "VarProp", "FSmart")
.ParameterOption(p => p.lowestPtTrack, "VarProp", "FSmart");
var mMVA = t.AddMethod(ROOTNET.Interface.NTMVA.NTypes.EMVA.kBDT, "SimpleBDT")
.Option("MaxDepth=3");
t.Train("VerySimpleTraining");
// Lets make a measurement of the efficiency for the standard cut.
var effResults = new FutureTFile(Path.Combine(FileInfoUtilities.FindDirectoryWithFileMatching("*.sln").FullName,"trainingResults.root"));
var stdCutVale = CalcEff(effResults.mkdir("std"), td => (td.logR > 1.2 && td.nTracks == 0) ? 1.0 : 0.0, 0.5, TrainingIQueriable(backgroundEvents, false), TrainingIQueriable(signalHV600pi100Events, true));
// Next, do it with a reader
#if false
var r = new ROOTNET.NTMVA.NReader();
var s = new FileInfo("C:\\Users\\gordo\\Documents\\Code\\calratio2015\\JetCutStudies\\SimpleJetCutTraining\\bin\\x86\\Debug\\weights\\VerySimpleTraining_SimpleCuts.weights.xml");
float[] logR = new float[2];
float[] nTracks = new float[2];
r.AddVariable("logR".AsTS(), logR);
r.AddVariable("nTracks".AsTS(), nTracks);
r.BookMVA("SimpleCuts".AsTS(), s.FullName.AsTS());
Expression<Func<TrainingData, double>> simpleCutsReader = tv => TMVAReaders.TMVASelectorSimpleCutsTest(r, tv.logR, tv.lowestPtTrack, stdCutVale.Value);
#else
Expression<Func<TrainingData, double>> simpleCutsReader = tv => TMVAReaders.TMVASelectorSimpleCuts(tv.logR, tv.nTracks, 0.72);
#endif
var simpleCutValue = CalcEff(effResults.mkdir("SimpleCuts"), simpleCutsReader, 0.5, TrainingIQueriable(backgroundEvents, false), TrainingIQueriable(signalHV600pi100Events, true));
// Next, we need to get the MVA and figure out the efficiency.
Expression<Func<TrainingData, double>> simpleMVAReader = tv => TMVAReaders.TMVASelectorSimpleBDT(tv.logR, tv.nTracks);
var simpleBDTValue = CalcEff(effResults.mkdir("SimpleBDT"), simpleMVAReader, 0.99999, TrainingIQueriable(backgroundEvents, false), TrainingIQueriable(signalHV600pi100Events, true));
// Write out everything
effResults.Write();
effResults.Close();
//Emit();
}
public void FutureFileEmptyInUsing()
{
var f = new FileInfo("FutureFileWriteHisto.root");
if (f.Exists)
{
f.Delete();
}
using (var ftf = new FutureTFile(f))
{
}
f.Refresh();
Assert.IsTrue(f.Exists);
}
public void FutureFileEmpty()
{
// Create an empty TFile, make sure that is good!
var f = new FileInfo("FutureFileEmpty.root");
if (f.Exists)
{
f.Delete();
}
var ftf = new FutureTFile(f);
ftf.Write();
ftf.Close();
f.Refresh();
Assert.IsTrue(f.Exists);
}
static void Main(string[] args)
{
var opt = ParseOptions<Options>(args);
var signalSources = SampleMetaData.AllSamplesWithTag("signal")
.Select(info => Tuple.Create(info.NickName, Files.GetSampleAsMetaData(info)));
using (var outputHistograms = new FutureTFile("LLPInvestigations.root"))
{
foreach (var s in signalSources)
{
FutureWriteLine(s.Item1);
ProcessSample(s.Item2, outputHistograms.mkdir(s.Item1));
}
}
FutureConsole.DumpToCout();
}
/// <summary>
/// Make generic plots of the signal or background
/// </summary>
/// <param name="args"></param>
/// <remarks>
/// TODO: Would studies of efficiencies here be better served by splitting into forward and central eta regions?
/// TODO: What is going on with the jetPT?
/// TODO: What is eta distribution of the jets that make it through, in particular with NTrack = 0?
/// It could be those are far forward and thus have no tracks.
/// TODO: Should any of these plots look at stuff in the way that Heather has (2D heat maps for cuts)?
/// </remarks>
static void Main(string[] args)
{
var opt = CommandLineUtils.ParseOptions<Options>(args);
Console.WriteLine("Finding the files");
// All the background samples have to be done first.
var backgroundSamples = SampleMetaData.AllSamplesWithTag("background")
.Select(info => Tuple.Create(Files.GetSampleAsMetaData(info), info.NickName))
.ToArray();
var backgroundEvents = Files.GetAllJetSamples().Select(e => e.Data);
// All the signal we are going to make plots of.
var signalSamples = SampleMetaData.AllSamplesWithTag("signal")
.Select(info => Tuple.Create(Files.GetSampleAsMetaData(info), info.NickName))
.ToArray();
// Output file
Console.WriteLine("Opening output file");
using (var outputHistograms = new FutureTFile("GenericPerformancePlots.root"))
{
// First, lets do a small individual thing for each individual background sample.
var bkgDir = outputHistograms.mkdir("background");
Console.WriteLine("Making background plots.");
foreach (var background in backgroundSamples)
{
BuildSuperJetInfo(background.Item1.Select(md => md.Data))
.PlotBasicDataPlots(bkgDir.mkdir(background.Item2), "all");
}
// Do a quick study for each signal sample, using all the backgrounds at once to make
// performance plots.
Console.WriteLine("Making the signal/background plots.");
foreach (var sample in signalSamples)
{
var status = PerSampleStudies(backgroundEvents, sample.Item1.Select(md => md.Data), outputHistograms.mkdir(sample.Item2));
DumpResults($"Sample {sample.Item2}:", status);
}
// Write out the histograms
outputHistograms.Write();
}
}
public void FutureFileWriteWithClose()
{
var f = new FileInfo("FutureFileWriteWithClose.root");
if (f.Exists)
{
f.Delete();
}
while (f.Exists)
{
f.Refresh();
}
using (var ftf = new FutureTFile(f))
{
ftf.Close();
}
f.Refresh();
Assert.IsTrue(f.Exists);
}
/// <summary>
/// Look at a number of MVA trainings and use them as results.
/// </summary>
/// <param name="args"></param>
static void Main(string[] args)
{
// Parse the arguments.
var opt = CommandLineUtils.ParseOptions<Options>(args);
// Get all the samples we want to look at, and turn them into
// jets with the proper weights attached for later use.
var backgroundJets = CommandLineUtils.GetRequestedBackground();
var allBackgrounds = new List<Tuple<string, IQueryable<Files.MetaData>>>()
{
Tuple.Create("QCD", backgroundJets),
};
var allSources = SampleMetaData.AllSamplesWithTag(opt.SignalTag)
.Select(info => Tuple.Create(info.NickName, Files.GetSampleAsMetaData(info, false)))
.ToArray();
if (allSources.Length == 0)
{
throw new ArgumentException($"No samples were found with tag '{opt.SignalTag}'.");
}
// List the artifacts that we are going to be using.
var mvaResults = new MVAInfo[]
{
new MVAInfo() { Name = "13VariableLxy", Artifact = new Uri("http://higgs.phys.washington.edu:8080/job/CalRatio2016/job/JetMVATraining/38/artifact/Jet.MVATraining-Jet.CalRat.NTrac.SumPtOfAllTrac.MaxTrack.JetWid.JetDRTo2GeVTra.EnergyDensi.HadronicLayer1Fracti.JetL.JetLo.FirstClusterRadi.ShowerCent_BDT.weights.xml") },
new MVAInfo() { Name = "13VariableLxyNoDR", Artifact = new Uri("http://higgs.phys.washington.edu:8080/job/CalRatio2016/job/JetMVATraining/39/artifact/Jet.MVATraining-Jet.CalRat.NTrac.SumPtOfAllTrac.MaxTrack.JetWid.EnergyDensi.HadronicLayer1Fracti.JetL.JetLo.FirstClusterRadi.ShowerCent_BDT.weights.xml") },
new MVAInfo() { Name = "13VariableLxyNoDRNTrack", Artifact = new Uri("http://higgs.phys.washington.edu:8080/job/CalRatio2016/job/JetMVATraining/40/artifact/Jet.MVATraining-Jet.CalRat.SumPtOfAllTrac.MaxTrack.JetWid.EnergyDensi.HadronicLayer1Fracti.JetL.JetLo.FirstClusterRadi.ShowerCent_BDT.weights.xml") },
new MVAInfo() { Name = "13VariableLxyNoNTrack", Artifact = new Uri("http://higgs.phys.washington.edu:8080/job/CalRatio2016/job/JetMVATraining/41/artifact/Jet.MVATraining-Jet.CalRat.NTrac.SumPtOfAllTrac.MaxTrack.JetWid.EnergyDensi.HadronicLayer1Fracti.JetL.JetLo.FirstClusterRadi.ShowerCent_BDT.weights.xml") },
};
// Fill an output file with the info for each MVA
using (var f = new FutureTFile("TrainingTestResults.root"))
{
foreach (var mva in mvaResults)
{
// The directory
var d = f.mkdir(mva.Name);
// Get the weight file.
var weights = ArtifactAccess.GetArtifactFile(mva.Artifact).Result;
// Now, create the MVA value from the weight file
var mvaValue = MVAWeightFileUtils.MVAFromWeightFile<TrainingTree>(weights);
// Do the backgrounds (e.g. ones where we don't filter for signal).
foreach (var s in allBackgrounds)
{
var sampleD = d.mkdir(s.Item1);
PlotMVAResult(s.Item2.AsGoodJetStream().FilterNonTrainingEvents(), sampleD, mvaValue);
}
// And now we can make the plots for signal
foreach (var s in allSources)
{
Console.WriteLine($"{mva.Name} - {s.Item1}");
var sampleD = d.mkdir(s.Item1);
PlotMVAResult(s.Item2.AsGoodJetStream().FilterNonTrainingEvents().FilterLLPNear(), sampleD, mvaValue);
}
}
}
}
static void Main(string[] args)
{
// Get command line arguments
var opt = ParseOptions<Options>(args);
// Build our own set of background samples so we can experiment.
var jets = SampleMetaData.AllSamplesWithTag("background")
.Select(info => Files.GetSampleAsMetaData(info));
// Count them individually
var individualCounts = jets.Select(sample => sample.FutureCount()).ToArray();
var firstsum = from c1 in individualCounts[0] from c2 in individualCounts[1] select c1 + c2;
var sum = individualCounts.Skip(2).Aggregate(firstsum, (tot, val) => from t in tot from v in val select t + v);
// Get the samples with an official weight attached to them.
var allSamplesToTest =
new[] { Tuple.Create("JZAll", GetAllJetSamples()) }
.Concat(SampleMetaData.AllSamplesWithTag("background").Select(info => Tuple.Create(info.NickName, Files.GetSampleAsMetaData(info))))
.ToArray();
var totalCount = GetAllJetSamples().FutureCount();
// Make a pT plot
using (var outputHistograms = new FutureTFile("JZPlotter.root"))
{
foreach (var sampleInfo in allSamplesToTest)
{
var events = sampleInfo.Item2;
var hdir = outputHistograms.mkdir(sampleInfo.Item1);
events
.Select(e => e.Data.eventWeight)
.FuturePlot("event_weights", "Sample EventWeights", 100, 0.0, 1000.0)
.Save(hdir);
events
.AsGoodJetStream()
.FuturePlot(JetPtPlotJetStream.ResetWeight(), "pt_unweighted")
.Save(hdir);
events
.AsGoodFirstJetStream()
.FuturePlot(JetPtPlotJetStream.ResetWeight(), "first_pt_unweighted")
.Save(hdir);
events
.AsGoodJetStream()
.FuturePlot(JetPtPlotJetStream, "pt_weighted")
.Save(hdir);
events
.AsGoodFirstJetStream()
.FuturePlot(JetPtPlotJetStream, "first_jet_pt_weighted")
.Save(hdir);
}
}
// Print out summary of numbers
Console.WriteLine($"Sum by adding individually: {sum.Value}.");
Console.WriteLine($"Sum when done all as one: {totalCount.Value}");
foreach (var s in individualCounts)
{
Console.WriteLine($" Sample Events: {s.Value}.");
}
}
