3D的博客

代码makegenerators.pyimport osimport tushare as tsimport numpy as npimport pandas as pddef makegenerators(ts_code='600004.SH', batch_size=1024, start_rate='', end_rate=''): # 判断文件是否存在,不存在则通...

论文Wasserstein Divergence for GANs提出了WGAN-div,主要是针对 1-Lipschitz问题提出了一种新的解决方案.个人感觉比WGAN-GP要好.以下代码的超参数是根据论文中的写的,但是网络架构并不是.论文中使用的是ResNet结构本文中用的是普通的卷积网络架构#! -*- coding: utf-8 -*-# wgan-divimport msv...

train_on_batch()返回的loss是执行这次训练前的loss使用相同的样本进行train和evaluate得到的loss不同主要是因为网络中的Dropout和BatchNormalization在训练模式/测试模式(learning_phase)中表现不同https://keras.io/getting-started/faq/#why-is-the-training-loss-m...

import osimport tensorflow as tfimport keras.backend.tensorflow_backend as KTFfrom keras.preprocessing import imagefrom keras.preprocessing.image import ImageDataGeneratorimport kerasfrom keras ...

文字感情分析参考:https://nbviewer.jupyter.org/github/fchollet/deep-learning-with-python-notebooks/blob/master/6.2-understanding-recurrent-neural-networks.ipynb# 数据预处理from keras.datasets import imdbfrom ke...

参考:https://nbviewer.jupyter.org/github/fchollet/deep-learning-with-python-notebooks/blob/master/6.1-using-word-embeddings.ipynb# ************* 加载数据import os# download from http://ai.stanford.edu/~a...

数据集文件结构代码import kerasfrom keras import layersfrom keras import modelsfrom keras.applications import VGG16from keras.preprocessing.image import ImageDataGeneratorinput_size = (150, 150)batch...

卷积核可视化import matplotlib.pyplot as pltimport numpy as npfrom keras import backend as Kfrom keras.models import load_model# 将浮点图像转换成有效图像def deprocess_image(x): # 对张量进行规范化 x -= x.mean() ...

训练过程def show_train_history(train_history, train_metrics, validation_metrics): plt.plot(train_history.history[train_metrics]) plt.plot(train_history.history[validation_metrics]) plt.title(...

目录数据来源文件结构代码load_tools.pyget_tools.pyget_samples.pynew_generator.pyevaluate_model.pyhistory_predict.pyserch_predict.pyrun.pyrun.batdotrain.py效果DNNCNNGRUResNet混合Attention总结数据来源使用tushare的接口获取股票历史数据.#...