MLA_Learn_Periodic_Functions/code/RNN_prediction_meteo/training.py

import tensorflow as tf
from matplotlib import pyplot as plt
import numpy as np
from Model import MyModel
import optuna

def objectif(trial) :
	LEN_SEQ = trial.suggest_int('len_seq', 64,256)
	PRED = 0.1
	HIDDEN = trial.suggest_int('hidden',32,256)

	model = MyModel(HIDDEN)

	dataset = np.load('dataset.npy')
	scale = (np.max(dataset) - np.min(dataset))
	data = dataset/scale
	shift = np.min(data)
	data = data - shift

	annee = np.array(list(range(len(data))))/365
	annee = annee - annee[-1]
	start_pred = int(len(data)*(1-PRED))
	#print(len(data))
	#print(start_pred)
	plt.figure(1)
	plt.plot(annee[:start_pred],data[:start_pred], label="apprentissage")
	plt.plot(annee[start_pred:],data[start_pred:], label="validation")
	plt.legend()
	#plt.show()

	X_train_tot = [data[0:start_pred-1]]
	Y_train_tot = [data[1:start_pred]]

	X_train_tot = np.expand_dims(np.array(X_train_tot),2)
	Y_train_tot = np.expand_dims(np.array(Y_train_tot),2)

	X_train = X_train_tot[:,:LEN_SEQ,:]
	Y_train = Y_train_tot[:,:LEN_SEQ,:]

	for i in range(len(X_train_tot[0]) - LEN_SEQ) :
		X_train = np.concatenate((X_train, X_train_tot[:,i:i+LEN_SEQ,:]),0)
		Y_train = np.concatenate((Y_train, Y_train_tot[:,i:i+LEN_SEQ,:]),0)
	#print(X_train_tot.shape)
	#print(X_train.shape)


	model.compile(optimizer='adam',
				  loss='binary_crossentropy',
				  metrics=['binary_crossentropy','mse'])
	import os 
	os.system("rm -rf log_dir")

	model.fit(x=X_train, y=Y_train, batch_size=16, epochs=5, shuffle=True)

	Pred = X_train_tot.copy()
	while len(Pred[0]) < len(data)+500 :
		print(len(data) - len(Pred[0]))
		Pred = np.concatenate((Pred, np.array([[model.predict(Pred)[0][-1]]])),1)
	Pred = Pred + shift
	Pred = Pred * scale

	data_Pred = np.squeeze(Pred)

	plt.figure(2)
	plt.clf()
	plt.plot(annee[:start_pred],dataset[:start_pred], label="apprentissage")
	plt.plot(annee[start_pred:],dataset[start_pred:], label="validation")
	plt.plot(annee[start_pred:], data_Pred[start_pred:len(annee)], label="prediction")
	plt.legend()
	plt.savefig(f"images/{LEN_SEQ}_{HIDDEN}_fig2.png")

	plt.figure(3)
	plt.clf()
	fig, axs = plt.subplots(2)
	axs[0].plot(annee[:start_pred],dataset[:start_pred], label="apprentissage")
	axs[0].plot(annee[start_pred:],dataset[start_pred:], label="validation")
	axs[1].plot(annee[:start_pred],dataset[:start_pred], label="données")
	axs[1].plot(annee[start_pred:], data_Pred[start_pred:len(annee)], label="prediction")
	plt.legend()
	plt.savefig(f"images/{LEN_SEQ}_{HIDDEN}_fig3.png")

	annee_ex = np.array(list(range(len(data_Pred))))/365
	annee_ex = annee_ex + annee[0]
	plt.figure(4)
	plt.clf()
	plt.plot(annee[:start_pred],dataset[:start_pred], label="données")
	plt.plot(annee_ex[start_pred:], data_Pred[start_pred:], label="prediction")
	plt.legend()
	plt.savefig(f"images/{LEN_SEQ}_{HIDDEN}_fig4.png")

	loss = np.linalg.norm(dataset[start_pred:] - data_Pred[start_pred:len(dataset)])/(np.max(data_Pred[start_pred:]) - np.min(data_Pred[start_pred:]))
	print(loss*100)
	return loss
	#plt.show()

study = optuna.create_study()
study.optimize(objectif, n_trials=10)