def evaluate_performance(self, class_labels, preds, gold, cl_perf = None, overall_perf = True, desc = " () ", print_perf = True): conf_matrix = confusion_matrix.ConfusionMatrix(class_labels, preds, gold, self.one_hot_encoding_preds, self.class_indices) if print_perf: if cl_perf is not None: for cl in cl_perf: p, r, f = conf_matrix.get_class_performance(cl) print(desc + " Class: " + cl + "\nP: " + str(p) + "\nR: " + str(r) + "\nF: " + str(f) + "\n") if overall_perf: print(desc + " Micro F1: " + str(conf_matrix.microf1) + "\nMacro F1: " + str(conf_matrix.macrof1) + "\n") return conf_matrix
def eval_func(golds, preds, params=None):
gold_labs = np.argmax(golds, axis=1)
pred_labs = np.argmax(preds, axis=1)
conf_matrix = confusion_matrix.ConfusionMatrix(params["dist_labels"],
pred_labs,
gold_labs,
False,
class_indices=True)
res = conf_matrix.accuracy
return 0 if math.isnan(res) else res
def get_performance_seqlab(golds,
preds,
labels,
print=True,
print_per_class=True):
golds_flat = []
preds_flat = []
for i in range(len(golds)):
for j in range(len(golds[i])):
if np.count_nonzero(golds[i][j]) > 0:
golds_flat.append(golds[i][j])
preds_flat.append(preds[i][j])
else:
break
confmat = confusion_matrix.ConfusionMatrix(labels,
preds_flat,
golds_flat,
one_hot_encoding=True)
confmat.print_results()
return confmat.accuracy