GAN用于收益模拟

如何解决GAN用于收益模拟

我是python的新手，尽管我已经了解了很多事情。我一直在阅读有关GAN及其在金融中用于模拟股票收益的知识，该收益比模拟carlo更为精确。我一直在努力编写一个有效的模型。代码下方

from keras.layers import Input,Dense
from keras.layers.advanced_activations import LeakyReLU
from keras.models import Sequential,Model
from keras.optimizers import Adam
from sklearn.preprocessing import MinMaxScaler
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import yfinance as yf

gm = yf.download('GM',period='MAX')['Adj Close']
ret = np.log(gm/gm.shift()).dropna()
length = 25
scaler = MinMaxScaler()
ret_scaled = scaler.fit_transform(ret[:,np.newaxis])

def create_samples(dataset):
    samples = []
    
    for i in range(len(dataset)-length):
        samples.append(dataset[i:(i+length)])
    
    return np.array(samples)

samples = create_samples(ret)

optimizer = Adam(0.0002,0.5)
noise_shape = (100,) 


def build_generator():
      
    generator = Sequential()
    generator.add(Dense(256,input_shape=noise_shape))
    generator.add(LeakyReLU(alpha=0.2))
    generator.add(Dense(512))
    generator.add(LeakyReLU(alpha=0.2))
    generator.add(Dense(1024))
    generator.add(LeakyReLU(alpha=0.2))
    generator.add(Dense(2048))
    generator.add(LeakyReLU(alpha=0.2))
    generator.add(Dense(25,activation='tanh'))
    generator.compile(loss='binary_crossentropy',optimizer=optimizer)

    return generator

def build_discriminator():

    discriminator = Sequential()
    discriminator.add(Dense(1024,input_dim=25))
    discriminator.add(LeakyReLU(alpha=0.2))
    discriminator.add(Dense(512))
    discriminator.add(LeakyReLU(alpha=0.2))
    discriminator.add(Dense(256))
    discriminator.add(LeakyReLU(alpha=0.2))
    discriminator.add(Dense(1,activation='sigmoid'))
    discriminator.compile(loss='binary_crossentropy',optimizer=optimizer)

    return discriminator

def build_gan(discriminator,generator):
    
    discriminator.trainable = False
    gan_input = Input(shape=(100,))
    x = generator(gan_input)
    gan_output = discriminator(x)
    gan = Model(inputs=gan_input,outputs=gan_output)
    gan.compile(loss='binary_crossentropy',optimizer=optimizer)
    
    return gan

def train(epochs=1,batch_size=10):

    X_train = samples
    batch_count = X_train.shape[0] / batch_size
    
    d_loss_logs_r = []
    d_loss_logs_f = []
    g_loss_logs = []
    
    generator = build_generator()
    discriminator = build_discriminator()
    gan = build_gan(discriminator,generator)
    
    for epoch in range(epochs):
        
        noise= np.random.normal(0,1,[batch_size,100])
        gen_imgs = generator.predict(noise)
        y_fake = np.zeros(batch_size)
        
        idx = np.random.randint(0,X_train.shape[0],batch_size)
        image_batch = X_train[idx]
        y_real = np.ones(batch_size)
        
        discriminator.trainable=True
        d_loss_real = discriminator.train_on_batch(image_batch,y_real)
        d_loss_fake = discriminator.train_on_batch(gen_imgs,y_fake)
        d_loss = 0.5 * np.add(d_loss_real,d_loss_fake) 
        
        noise= np.random.normal(0,100])
        y_gen = np.ones(batch_size)
        discriminator.trainable=False
        g_loss = gan.train_on_batch(noise,y_gen)
        
        d_loss_logs_r.append([epoch,d_loss_real])
        d_loss_logs_f.append([epoch,d_loss_fake])
        g_loss_logs.append([epoch,g_loss])
        
    d_loss_logs_r_a = np.array(d_loss_logs_r)
    d_loss_logs_f_a = np.array(d_loss_logs_f)
    g_loss_logs_a = np.array(g_loss_logs)
    
    return d_loss_logs_r_a,d_loss_logs_f_a,g_loss_logs_a

当我训练模型时，它只产生鉴别器和生成器的损失函数，但是我又如何获得模拟收益（因此我可以使用它们代替蒙特卡洛的收益）？

非常感谢