I came across this page. It defines METRICS
as below. My questions are
METRICS = [
keras.metrics.TruePositives(name='tp'),
keras.metrics.FalsePositives(name='fp'),
keras.metrics.TrueNegatives(name='tn'),
keras.metrics.FalseNegatives(name='fn'),
keras.metrics.BinaryAccuracy(name='accuracy'),
keras.metrics.Precision(name='precision'),
keras.metrics.Recall(name='recall'),
keras.metrics.AUC(name='auc'),
]
Train on 182276 samples, validate on 45569 samples
Epoch 1/100
182276/182276 [==============================] - 2s 12us/sample - loss: 0.0139 - tp: 7.0000 - fp: 124.0000 - tn: 181835.0000 - fn: 310.0000 - accuracy: 0.9976 - precision: 0.0534 - recall: 0.0221 - auc: 0.7262 - val_loss: 0.0074 - val_tp: 4.0000 - val_fp: 0.0000e+00 - val_tn: 45492.0000 - val_fn: 73.0000 - val_accuracy: 0.9984 - val_precision: 1.0000 - val_recall: 0.0519 - val_auc: 0.8742
Epoch 2/100
182276/182276 [==============================] - 0s 3us/sample - loss: 0.0076 - tp: 91.0000 - fp: 30.0000 - tn: 181929.0000 - fn: 226.0000 - accuracy: 0.9986 - precision: 0.7521 - recall: 0.2871 - auc: 0.8828 - val_loss: 0.0053 - val_tp: 39.0000 - val_fp: 7.0000 - val_tn: 45485.0000 - val_fn: 38.0000 - val_accuracy: 0.9990 - val_precision: 0.8478 - val_recall: 0.5065 - val_auc: 0.8761
Epoch 3/100
182276/182276 [==============================] - 0s 3us/sample - loss: 0.0064 - tp: 146.0000 - fp: 36.0000 - tn: 181923.0000 - fn: 171.0000 - accuracy: 0.9989 - precision: 0.8022 - recall: 0.4606 - auc: 0.8981 - val_loss: 0.0049 - val_tp: 45.0000 - val_fp: 7.0000 - val_tn: 45485.0000 - val_fn: 32.0000 - val_accuracy: 0.9991 - val_precision: 0.8654 - val_recall: 0.5844 - val_auc: 0.8828
METRICS
. Is loss
a default option? would it be present for regression or multi class classification too?METRICS
for training and validation data. Is it because when we fit the model we provide validation data validation_data=(val_features, val_labels)
? if we dont provide validation data, would that give an error as it couldnt print metrics for validation data?1 - It is default, unless you put verbose=0
then there is nothing. Yes in all case it is present
2 - Yes, if you don't provide validation_data then you won't have anything on val metrics