Example Python code for performing multi-output regression with Keras and TensorFlow

multi_output_regression_keras.py

# Download the 'train_embeddings.pkl' from https://files.fm/f/76h5d4229
# Run the code in the folder where 'train_embeddings.pkl' is located.

import io

import compress_pickle as pickle
import numpy as np
import tensorflow as tf
from keras import Sequential
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
from keras.optimizers import Adam
from matplotlib import pyplot as plt
from sklearn.model_selection import train_test_split

# Model parameters
learning_rate = 1e-3
epochs = 20
patience = 3
batch_size = 32

with io.open("train_embeddings.pkl", "rb") as inf:
    train_data = pickle.load(inf)

X = np.array(train_data["embedding"])
# y = np.array([[i[0]] for i in train_data["labels"]])
y = np.array(train_data["labels"])

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, random_state=60)

print(X_train.shape, y_train.shape)
print(X_test.shape, y_test.shape)

model = Sequential([
    tf.keras.layers.Dense(20, input_dim=X_train[0].shape[0], kernel_initializer='he_uniform', activation="relu"),
    tf.keras.layers.Dense(128, activation="relu"),
    tf.keras.layers.Dense(128, activation="relu"),
    tf.keras.layers.Dense(y[0].shape[0], activation=None)
])

callbacks = [ModelCheckpoint(filepath="best_model.keras", save_best_only=True, monitor="val_loss"),
             ReduceLROnPlateau(monitor='val_loss', factor=1e-1, patience=patience, min_lr=1e-5),
             EarlyStopping(monitor='val_loss', patience=patience)]

optimizer = Adam(learning_rate=learning_rate)
# optimizer = RMSprop(learning_rate=learning_rate)
# optimizer = SGD(learning_rate=learning_rate)

model.compile(loss="mae", optimizer=optimizer)

print(model.summary())

history = model.fit(x=X_train, y=y_train, validation_data=(X_test, y_test), epochs=epochs, callbacks=callbacks,
                    batch_size=batch_size)

loss = history.history["loss"]
val_loss = history.history["val_loss"]

epochs_ = range(1, len(loss) + 1)
plt.figure()
plt.plot(epochs_, loss, "b", label="Training loss")
plt.plot(epochs_, val_loss, "r", label="Validation loss")
plt.title("Training and validation loss")
plt.xlabel("epoch")
plt.ylabel("loss")
plt.legend()

plt.show()