public static Dataset KeepAboveConfidenceThreshold(NeuralNet net, Dataset ds, double conf)
{
return(Filter(net, ds, (nn, datum, ground_label) => {
LabelWithConfidence lab = Utils.ULabel.LabelWithConfidence(nn, datum, true);
return (lab.softMaxValue >= conf);
}));
}
public static LabelWithConfidence LabelWithConfidence(NeuralNet model, NNInstrumentation instr, double[] datum, bool crop)
{
Vector <double> datum_v = DenseVector.OfArray(datum);
if (crop)
{
datum_v = model.CropMaybe(datum_v);
}
double[] outs = model.EvaluateNNConcretePostCrop(datum_v, instr);
// Console.WriteLine("Outs = {0}", DenseVector.OfArray(outs));
Tuple <double, int> max = UMath.Max(outs);
Tuple <double, int> secmax = UMath.MaxExcluding(max.Item2, outs);
UMath.SoftMax(outs);
var result = new LabelWithConfidence
{
datum = datum,
actualLabel = max.Item2,
secBestLabel = secmax.Item2,
softMaxValue = outs[max.Item2],
diffFromSecondBest = Math.Abs(outs[max.Item2] - outs[secmax.Item2])
};
return(result);
}
public static Dataset KeepMisclass(NeuralNet net, Dataset ds)
{
return(Filter(net, ds, (nn, datum, ground_label) =>
{
LabelWithConfidence lab = Utils.ULabel.LabelWithConfidence(nn, datum, true);
return (lab.actualLabel != ground_label);
}));
}
public static LabelWithConfidence[] LabelWithConfidence(NeuralNet model, Dataset input)
{
LabelWithConfidence[] result = new LabelWithConfidence[input.Count()];
for (int i = 0; i < input.Count(); i++)
{
result[i] = LabelWithConfidence(model, input.GetDatum(i), true);
}
return(result);
}
public void RecordAtomically(
String datasetName,
int datasetIndex,
LabelWithConfidence origLab,
LabelWithConfidence synthLab,
double scale, double offset,
int numRows, int numCols, bool isColor, bool isRowOrder = true)
{
SynthEntry e = new SynthEntry();
e.datasetName = datasetName;
e.origPngName = CreatePnG("orig", datasetName, datasetIndex, origLab.datum, scale, offset, numRows, numCols, isColor, isRowOrder);
e.origLabel = origLab.actualLabel;
e.synthPngName = CreatePnG("snth", datasetName, datasetIndex, synthLab.datum, scale, offset, numRows, numCols, isColor, isRowOrder);
e.synthLabel = synthLab.actualLabel;
double[] diff = new double[origLab.datum.Length];
for (int i = 0; i < origLab.datum.Length; i++)
{
diff[i] = 5 * (origLab.datum[i] - synthLab.datum[i]) + 100; // -20 ... 20 -> -100 .. 100 -> 0 .. 200
}
CreatePnG("diffx5o100", datasetName, datasetIndex, diff, scale, offset, numRows, numCols, isColor, isRowOrder);
Console.WriteLine("Orig path =" + e.origPngName);
Console.WriteLine("Synth path =" + e.synthPngName);
e.lInfDist = Utils.UMath.LInfinityDistance(origLab.datum, synthLab.datum);
e.l1Dist = Utils.UMath.L1Distance(origLab.datum, synthLab.datum);
e.confOrig = origLab.softMaxValue;
e.confSynth = synthLab.softMaxValue;
e.sndBestDiffOrig = origLab.diffFromSecondBest;
e.sndBestDiffSynth = synthLab.diffFromSecondBest;
SerializeEntry(e);
}
/// <summary>
/// Synthesize a counterexample from an existing labelled image.
/// </summary>
/// <param name="options"></param>
/// <param name="nn">The model.</param>
/// <param name="imageLab">The image and labeling information from the network.</param>
/// <param name="instr"></param>
/// <param name="realLabel">The label of the image from the training set.</param>
/// <param name="rowSize"></param>
/// <param name="colSize"></param>
/// <param name="isColor"></param>
/// <returns>NULL if we were not able to synthesize a counterexample, otherwise some information about it.</returns>
public static Nullable<LabelWithConfidence> SynthesizeCounterexample
( NeuralNet nn
, LPSTerm[] inputs // Symbolic inputs (cropped)
, LPSTerm epsilon // Epsilon variable
, LabelWithConfidence imageLab // Original image classification info (uncropped)
, NNInstrumentation instr
, int realLabel // Ground truth for this image (from training set)
, int rowSize // Original (uncropped) row size
, int colSize // Original (uncropped) col size
, bool isColor)
{
int origLabel = imageLab.actualLabel;
int targetLabel = imageLab.secBestLabel;
int input_dimension_pre_crop = nn.InputDimensionPreCrop;
int input_dimension_post_crop = nn.InputDimensionPostCrop;
double[] orig_image = imageLab.datum;
double[] orig_image_crop = nn.CropMaybe(DenseVector.OfArray(orig_image)).ToArray();
if (realLabel != origLabel)
{
Console.WriteLine("This image is misclassifed already! Skipping.");
return null;
}
if (RobustnessOptions.IgnoreLowConfidence && imageLab.softMaxValue < RobustnessOptions.LowConfidenceThreshold)
{
Console.WriteLine("This image is misclassifed with low confidence! Skipping.");
return null;
}
// Fast path:
// DiffInfo diff_info;
/* *********************
* DV: Commenting out the fast path for now (but we are still keeping the Dictionary, for debugging)
* *********************
if (diffDict.TryGetValue(new Tuple<int,int>(origLabel,targetLabel),out diff_info))
{
Console.WriteLine("Got a hit in the difference cache!");
Vector<double> diff_counterexample = diff_info.diff;
Vector<double> cand = DenseVector.OfArray(orig_image) + diff_counterexample;
Console.WriteLine("oooooooooooooooo Checking with the fast path!");
double[] cand_arr_crop = nn.CropMaybe(cand).ToArray();
if (RobustnessOptions.QuantizationSafety)
{
Utils.UArray.InPlaceRoundDoubleArray(cand_arr_crop);
}
LabelWithConfidence candLab = Utils.ULabel.LabelWithConfidence(nn, cand_arr_crop,false); // Already cropped, don't crop!
if (candLab.actualLabel != origLabel)
{
Console.WriteLine("=> Real counterexample (from fast path)!");
diff_info.number++;
return candLab;
}
Console.WriteLine("xxxx Fast path failed, continuing with symbolic interpreter ...");
// otherwise continue with the slow path ...
}
***********************/
var state = new LPSState(instr, orig_image_crop);
int nomodelcount = 0;
double[] newImageUnrounded;
NOMODELLOOP:
if (nomodelcount++ > 0) return null;
state.ClearConstraints();
LPSTerm[] output = nn.EvaluateNNSymbolicPostCrop(state, inputs);
// Just some tracing ...
// ReportSparsity(output);
LPSConstraints currentCts = state.CurrentCts;
LPSConstraints deferredCts = state.DeferredCts;
// Conjoin the label formula
currentCts.And(NNetFormulas.LabelFormula(output, targetLabel, RobustnessOptions.LabelConfidenceDiff));
// If we are just looking for bounds, then the variables themselves will contain "origin" bounds
if (RobustnessOptions.DoOptimization)
{
NNETObjectives.AddEpsilonBounds(currentCts, inputs, epsilon, orig_image_crop);
}
// Ensure that at least *one* entry is different by at least 1.0
if (RobustnessOptions.QuantizationSafety)
{
NNETObjectives.AddQuantizationSafety(currentCts, inputs, orig_image_crop);
}
// Create objective
Nullable<LPSObjective> objective = null;
if (RobustnessOptions.DoOptimization)
{
switch (RobustnessOptions.ObjectiveKind)
{
case LPSObjectiveKind.MinLinf:
objective = NNETObjectives.MinLInf(currentCts, inputs, epsilon, orig_image_crop);
break;
case LPSObjectiveKind.MaxConf:
objective = NNETObjectives.MaxConf(output, origLabel, targetLabel);
break;
default:
break;
}
}
if (!RobustnessOptions.CEGAR)
{
currentCts.And(deferredCts);
deferredCts = new LPSConstraints();
}
// CEGAR loop header
LabelWithConfidence newLab;
Console.WriteLine(
"Current constraints: {0}, deferred: {1}",
currentCts.Count,
deferredCts.Count);
LPSolver lps = new LPSolver(
input_dimension_post_crop,
currentCts.Count + deferredCts.Count,
orig_image_crop,
RobustnessOptions.Epsilon);
lps.AddConstraints(currentCts, objective);
int cegar_iterations = 0;
while (true)
{
if (cegar_iterations++ > RobustnessOptions.CEGARGiveUpIterations)
{
Console.WriteLine("xxxxxxxxxxxxxxxx Giving up CEGAR, could not find model!");
goto NOMODELLOOP;
}
var newImage = lps.SolveLowLevelLP();
currentCts = new LPSConstraints();
if (newImage == null)
{
Console.WriteLine("xxxxxxxxxxxxxxxx No model!");
goto NOMODELLOOP;
}
Console.WriteLine("oooooooooooooooo Found model!");
newImageUnrounded = new double[newImage.Length];
Array.Copy(newImage, newImageUnrounded, newImage.Length);
if (RobustnessOptions.QuantizationSafety)
{
Utils.UArray.InPlaceRoundDoubleArray(newImage);
}
int samcount = Utils.UArray.ComputeRoundIdenticals(orig_image_crop, newImage);
Console.WriteLine("Synthesized image has {0} identical inputs (after rounding) to original (cropped)", samcount);
// Now, try to label the new example
newLab = Utils.ULabel.LabelWithConfidence(nn, newImage,false); // Already cropped, don't crop!
if (newLab.actualLabel != targetLabel)
{
if (newLab.actualLabel == realLabel)
{
// Here the synthesized image is not really a counterexample.
// This could be due to either (a) quantization errors or (b) CEGAR
// underapproximation. But the only thing we can try and do here is
// add mor constraints and try to resolve.
if (RobustnessOptions.CEGAR)
Console.WriteLine("Not really a counterexample, going round CEGAR loop.");
int added = 0;
// new_image_plus_eps = newImage : 0.0
// so that the length matches the coefficients of each constraint ...
double[] newimage_plus_eps = new double[newImage.Length+1];
Array.Copy(newImageUnrounded,newimage_plus_eps,newImage.Length);
newimage_plus_eps[newImage.Length] = 0.0;
Vector<double> newImageVec_eps = DenseVector.OfArray(newimage_plus_eps);
var denumerator = deferredCts.GetEnumerator();
Parallel.For(0, deferredCts.Count, i =>
{
LPSConstraint curr_deferred;
if (added > 699) return;
lock (lockObj)
{
denumerator.MoveNext();
curr_deferred = (LPSConstraint)denumerator.Current;
if (curr_deferred.Added == true) return;
}
bool sat = Satisfiable(curr_deferred, newImageVec_eps);
lock (lockObj)
{
if (!sat)
{
lps.AddConstraint(curr_deferred);
// currentCts.And(curr_deferred.Term, curr_deferred.Inequality);
curr_deferred.Added = true;
added++;
}
}
});
Console.WriteLine();
Console.WriteLine("Added {0} constraints for CEGAR", added);
if (added == 0)
{
Console.WriteLine("=> CEGAR cannot improve things.");
goto NOMODELLOOP;
// return null;
}
// lps.AddConstraints(currentCts, null);
continue;
}
else
{
Console.WriteLine("=> Real counterexample! (Although with different label than expected)");
break;
}
}
else
{
Console.WriteLine("=> Real counterexample! (New image has second-best label");
break;
}
}
if (RobustnessOptions.DisplaySynthesizedImagesAndPause)
{
Utils.UDraw.DisplayImageAndPause(Utils.UArray.ToIntArray(imageLab.datum), rowSize, colSize, isColor);
Utils.UDraw.DisplayImageAndPause(Utils.UArray.ToIntArray(newLab.datum), rowSize, colSize, isColor);
}
/* NB: Uncrop the image in newLab */
newLab.datum = nn.UnCropMaybe(DenseVector.OfArray(orig_image), DenseVector.OfArray(newLab.datum)).ToArray();
double[] tmp = nn.UnCropMaybe(DenseVector.OfArray(orig_image), DenseVector.OfArray(newImageUnrounded)).ToArray();
Vector<double> diff_val = DenseVector.OfArray(tmp) - DenseVector.OfArray(orig_image);
var key = new Tuple<int, int>(origLabel, newLab.actualLabel);
DiffInfo dinfo;
if (diffDict.TryGetValue(key, out dinfo))
{
dinfo.number++;
}
else
{
dinfo = new DiffInfo();
dinfo.diff = diff_val;
dinfo.number = 1;
diffDict.Add(new Tuple<int, int>(origLabel, newLab.actualLabel), dinfo);
}
return newLab;
}
/// <summary>
/// Generate and return a list of counterexamples by iterating over the training set
/// </summary>
/// <param name="datasetname"></param>
/// <param name="options"></param>
/// <param name="nn"></param>
/// <param name="ds"></param>
/// <returns></returns>
public static List<LabelWithConfidence> SynthesizeCounterexamples(
NeuralNet nn,
ImageDataset ds,
Action<LabelWithConfidence> snapshot)
{
/* Initialization stuff */
List<LabelWithConfidence> counterexamples = new List<LabelWithConfidence>();
SynthRegistry reg = new SynthRegistry(RobustnessOptions.Registry + ".csv", RobustnessOptions.Registry);
/* How many training points to do */
int trainingPointsToDo = (int)Math.Round(ds.Dataset.Count() * RobustnessOptions.DataSetPercentage);
int completed = 0;
/* The symbolic variables: NB we use the dimension PostCrop to avoid generating lots of useless variables */
Tuple<LPSTerm[],LPSTerm> inputs = GenSymbolicInputs(nn.InputDimensionPostCrop);
// Alternatively (the code is thread-safe already):
// Parallel.For(0, ds.Dataset.Count(), RobustnessOptions.ParallelOptions, i =>
for (int i = 0; i < ds.Dataset.Count(); i++)
{
if (completed < trainingPointsToDo)
{
Console.WriteLine("Image count = {0}", i);
NNInstrumentation instr = new NNInstrumentation();
LabelWithConfidence imageLab = ULabel.LabelWithConfidence(nn, instr, ds.Dataset.GetDatum(i), true);
Nullable<LabelWithConfidence> synthLab = null;
try
{
var stopwatch = new Stopwatch();
stopwatch.Start();
synthLab = SynthesizeCounterexample(
nn,
inputs.Item1,
inputs.Item2,
imageLab,
instr,
ds.Dataset.GetLabel(i),
ds.RowCount,
ds.ColumnCount,
ds.IsColor);
stopwatch.Stop();
Console.WriteLine("Processed image in {0} milliseconds", stopwatch.ElapsedMilliseconds);
GC.Collect();
}
catch
{
lock(lockObj)
{
completed++;
}
continue;
}
lock (lockObj)
{
completed++;
if (synthLab.HasValue)
{
// VERY IMPORTANTLY: Change the label of the counterexample
// to be the label of the original point! This was a horrible bug.
var forRetraining = synthLab.Value;
forRetraining.actualLabel = imageLab.actualLabel;
counterexamples.Add(forRetraining);
snapshot(forRetraining);
Console.WriteLine("forRetraining.label = {0}", forRetraining.actualLabel);
Console.WriteLine("synthLab.Value.label = {0}", synthLab.Value.actualLabel);
reg.RecordAtomically(
RobustnessOptions.Registry,
i,
imageLab,
synthLab.Value,
RobustnessOptions.ScalePreProcessed,
RobustnessOptions.OffsetPreProcessed,
ds.RowCount,
ds.ColumnCount,
ds.IsColor);
}
Console.WriteLine("Counterexamples/Processed-so-far: " + counterexamples.Count() + "/" + completed);
}
}
}
return counterexamples;
}