nn.py

from ml_hpx import Layer, SGD, NeuralNetwork
import numpy as np
from time import perf_counter
import tensorflow as tf
from tensorflow.keras import Sequential
from tensorflow.keras.layers import Dense, Input

# ----- Dataset -----
X = np.array([[0,0], [0,1], [1,0], [1,1]], dtype=np.float32)
y = np.array([[0], [1], [1], [0]], dtype=np.float32)

# ----- HPX NN -----
layers = [
    Layer(4, 2, "relu"),
    Layer(1, 4, "sigmoid")
]
optimizer = SGD(0.1)
nn = NeuralNetwork(layers, optimizer)

start = perf_counter()
nn.fit(X.tolist(), y.tolist(), 1000)  # 5000 epochs
end = perf_counter()
print(f"HPX NN training time: {end - start:.6f} sec")

preds = nn.predict(X.tolist())
print("HPX Predictions:", preds)

# ----- TensorFlow NN -----
model = Sequential([
    Input(shape=(2,)),   # explicitly declare input shape
    Dense(4, activation='relu'),
    Dense(1, activation='sigmoid')
])

model.compile(optimizer=tf.keras.optimizers.SGD(learning_rate=0.1),
              loss='binary_crossentropy',
              metrics=['accuracy'])

start = perf_counter()
model.fit(X, y, epochs=1000, verbose=0)
end = perf_counter()
print(f"TensorFlow training time: {end - start:.6f} sec")

preds_tf = model.predict(X)
print("TF Predictions:", preds_tf.tolist())