95 lines
3.5 KiB
Python
95 lines
3.5 KiB
Python
|
import numpy as np
|
||
|
|
import tensorflow as tf
|
||
|
|
from tensorflow import keras
|
||
|
|
from tensorflow.keras.models import Sequential, Model
|
||
|
|
from tensorflow.keras.layers import Dense, SimpleRNN, Input, Add
|
||
|
|
import matplotlib.pyplot as plt
|
||
|
|
from sklearn.metrics import r2_score
|
||
|
|
|
||
|
|
|
||
|
|
def generate_data(n_samples):
|
||
|
|
X = np.random.uniform(-5, 5, (n_samples, 2))
|
||
|
|
Y = X[:, 0] * np.sin(X[:, 1]) + X[:, 1] * np.cos(X[:, 0])
|
||
|
|
return X, Y
|
||
|
|
|
||
|
|
# @tf.function()
|
||
|
|
def train_and_evaluate(model, X_train, Y_train, X_test, Y_test, epochs=100, batch_size=10):
|
||
|
|
model.compile(optimizer='adam', loss='mse')
|
||
|
|
history = model.fit(X_train, Y_train, epochs=epochs, batch_size=batch_size, verbose=0, validation_data=(X_test, Y_test))
|
||
|
|
Y_pred = model.predict(X_test)
|
||
|
|
error = np.mean(np.abs((Y_test - Y_pred) / Y_test))
|
||
|
|
r2 = r2_score(Y_test, Y_pred)
|
||
|
|
return history, error, Y_pred, r2
|
||
|
|
|
||
|
|
X, Y = generate_data(1000)
|
||
|
|
X_train, X_test = X[:800], X[800:]
|
||
|
|
Y_train, Y_test = Y[:800], Y[800:]
|
||
|
|
|
||
|
|
inputs = Input(shape=(2,))
|
||
|
|
hidden1 = Dense(20, activation='relu')(inputs)
|
||
|
|
output = Dense(1)(hidden1)
|
||
|
|
output_cascade = Dense(1)(inputs)
|
||
|
|
final_output = Add()([output, output_cascade])
|
||
|
|
cascade_model_1 = Model(inputs=inputs, outputs=final_output)
|
||
|
|
|
||
|
|
inputs_2 = Input(shape=(2,))
|
||
|
|
hidden1_2 = Dense(10, activation='relu')(inputs_2)
|
||
|
|
hidden2_2 = Dense(10, activation='relu')(hidden1_2)
|
||
|
|
output_2 = Dense(1)(hidden2_2)
|
||
|
|
output_cascade_2 = Dense(1)(inputs_2)
|
||
|
|
final_output_2 = Add()([output_2, output_cascade_2])
|
||
|
|
cascade_model_2 = Model(inputs=inputs_2, outputs=final_output_2)
|
||
|
|
|
||
|
|
models = {
|
||
|
|
"FeedForward (10 нейронів)": Sequential([Input(shape=(2,)), Dense(10, activation='relu'), Dense(1)]),
|
||
|
|
"FeedForward (20 нейронів)": Sequential([Input(shape=(2,)), Dense(20, activation='relu'), Dense(1)]),
|
||
|
|
"Cascade (20 нейронів)": cascade_model_1,
|
||
|
|
"Cascade (2x10 нейронів)": cascade_model_2,
|
||
|
|
"Elman (15 нейронів)": Sequential([Input((2, 1)), SimpleRNN(15, activation='relu'), Dense(1)]),
|
||
|
|
"Elman (3x5 нейронів)": Sequential([
|
||
|
|
Input((2, 1)), SimpleRNN(5, activation='relu', return_sequences=True),
|
||
|
|
SimpleRNN(5, activation='relu', return_sequences=True),
|
||
|
|
SimpleRNN(5, activation='relu'), Dense(1)
|
||
|
|
])
|
||
|
|
}
|
||
|
|
|
||
|
|
errors = {}
|
||
|
|
predictions = {}
|
||
|
|
r2_scores = {}
|
||
|
|
plt.figure(figsize=(10, 6))
|
||
|
|
|
||
|
|
for name, model in models.items():
|
||
|
|
print(f"Навчання {name}...")
|
||
|
|
if "Elman" in name:
|
||
|
|
X_train_rnn = X_train.reshape((X_train.shape[0], X_train.shape[1], 1) )
|
||
|
|
X_test_rnn = X_test.reshape((X_test.shape[0], X_test.shape[1], 1))
|
||
|
|
history, error, Y_pred, r2 = train_and_evaluate(model, X_train_rnn, Y_train, X_test_rnn, Y_test)
|
||
|
|
else:
|
||
|
|
history, error, Y_pred, r2 = train_and_evaluate(model, X_train, Y_train, X_test, Y_test)
|
||
|
|
|
||
|
|
errors[name] = error
|
||
|
|
predictions[name] = Y_pred
|
||
|
|
r2_scores[name] = r2
|
||
|
|
plt.plot(history.history['loss'], label=name)
|
||
|
|
|
||
|
|
plt.title("Зміна MSE під час навчання")
|
||
|
|
plt.xlabel("Епохи")
|
||
|
|
plt.ylabel("MSE")
|
||
|
|
plt.legend()
|
||
|
|
plt.show()
|
||
|
|
|
||
|
|
plt.figure(figsize=(10, 5))
|
||
|
|
for name, Y_pred in predictions.items():
|
||
|
|
plt.scatter(Y_test, Y_pred, label=name, alpha=0.5)
|
||
|
|
plt.plot(Y_test, Y_test, 'k--', label="Ідеальна лінія")
|
||
|
|
plt.xlabel("Реальні значення")
|
||
|
|
plt.ylabel("Передбачені значення")
|
||
|
|
plt.legend()
|
||
|
|
plt.title("Порівняння передбачень різних моделей")
|
||
|
|
plt.show()
|
||
|
|
|
||
|
|
for name, error in errors.items():
|
||
|
|
print(f"{name}: середня відносна помилка = {error:.4f}, R² = {r2_scores[name]:.4f}")
|
||
|
|
|
||
|
|
|