def objectHistogramMatrix(dataSet, features, exemplars, bins):
tasks = [(dataSet, features, exemplars, xFtr, yFtr, bins) for yFtr in features for xFtr in features if xFtr < yFtr]
if data.mdsColumnsPresent(dataSet):
tasks.append((dataSet, features, exemplars, data.mdsColumns[0], data.mdsColumns[1], bins))
pool = Pool()
results = pool.imap(objectHistogram2D, tasks)
pool.close()
pool.join()
recDict = lambda: defaultdict(recDict)
histograms = recDict()
for xFeature, yFeature, contours in results:
for c, lvlCnt in contours.iteritems():
histograms[xFeature][yFeature][c] = lvlCnt.tolist()
histograms[yFeature][xFeature][c] = lvlCnt.transpose().tolist()
return dict(histograms)
def objectInfo(dataSet, featureSet, column, row, plate, exemplars, probes):
objects = allObjects(dataSet, column, row, plate, exemplars, probes)
combined = wellIndex(dataSet).loc[objects].copy()
# Feature values.
for ftr in featureSet:
combined[ftr] = np.take(scaledArray(dataSet, ftr), objects)
if data.mdsColumnsPresent(dataSet):
for mdsCol in data.mdsColumns:
combined[mdsCol] = np.take(scaledArray(dataSet, mdsCol), objects)
# Predicted population values.
combined["population"] = np.take(clusters(dataSet, featureSet, exemplars), objects)
# Generate well URLs on the spot, based on config.
wellImages = data.config(dataSet).wellImages
for name, urlFunction in wellImages.iteritems():
combined["img_" + name] = combined.apply(
lambda row: urlFunction(int(row['plate']), int(row['column']), int(row['row'])), axis=1)
return combined