Some Metric Implementation in Keras (Such as Pearsons Correlation Coefficient, MRE)

Now Including:

• Pearsons Correlation Coefficient
• Mean Relative Error
• Jaccard Loss (Derivable, can be used as LOSS for training in Keras)
• Jaccard Index
• Dice Similarity Coefficient (aka. DSC)

Just an example ~

inp = Input(shape=(timesteps, 1))
gru = Bidirectional(GRU(500, return_sequences=True))(inp)

max1 = GlobalMaxPool1D()(gru_1)
att1 = Attention()(gru_1)
cont1 = keras.layers.concatenate([max1, att1])

out = Dense(1, activation='relu')(cont1)

opt = keras.optimizers.Adam(lr=0.002, beta_1=0.9, beta_2=0.9, epsilon=1e-08, amsgrad=True)

model = Model(inputs=inp, outputs=out)
model.compile(loss='mse',
              optimizer=opt,
              metrics=['mse', pearson_r, mre])

• Pearsons Correlation Coefficient

• Jaccard Loss

• Dice Similarity Coefficient

• Mean Relative Error

See here

Sayin C, Ertunc H M, Hosoz M, et al. Performance and exhaust emissions of a gasoline engine using artificial neural network[J]. Applied Thermal Engineering, 2007, 27(1):46-54.