weixin_42681868的博客

import tensorflow as tfimport matplotlib as mlpimport matplotlib.pyplot as plt%matplotlib inlineimport numpy as npimport pandas as pdimport sklearnimport osimport sysfrom tensorflow import ke...

import tensorflow as tfimport matplotlib as mlpimport matplotlib.pyplot as plt%matplotlib inlineimport numpy as npimport pandas as pdimport sklearnimport osimport sysfrom tensorflow import ke...

超参数搜索为什么要超参数搜索*神经网络有很多训练过程中不变的参数*网络结构参数:几层, 每层宽度(神经元个数), 每层的激活函数*训练参数: batch_size , 学习率 , 学习率衰减算法等手工去试耗费人力网格搜索遗传算法*对自然界的模拟*A:初始化候选参数结合 -> 训练 -> 得到的模型指标作为生存概率*B:选择 -> 交叉 -> 变异 -&g...

import matplotlib as mplimport matplotlib.pyplot as plt%matplotlib inlineimport numpy as npimport sklearnimport pandas as pdimport osimport sysimport tensorflow as tffrom tensorflow import ke...

稀疏特征和密集特征:稀疏特征:* 离散值特征* One-hot* 叉乘~ 稀疏特征做叉乘获取共现信息~ 实现记忆效果稀疏特征的优缺点:* 优点~ 对重复样板高校,广泛应用与工业界* 缺点~ 需要人工设计~ 可能会过拟合密集特征:* 向量表达~ 如 [0.1,0.2,0.6]*word2vec密集特征的优缺点* 优点~ 带有语义信息,不同向量之间有相关性~ 兼...

import matplotlib as mplimport matplotlib.pyplot as plt%matplotlib inlineimport numpy as npimport sklearnimport pandas as pdimport osimport sysimport tensorflow as tffrom tensorflow import ke...

import tensorflow as tffrom tensorflow import kerasimport matplotlib as mplfrom matplotlib import pyplot as plt%matplotlib inlineimport sklearnimport numpy as npimport pandas as pdimport osim...

import tensorflow as tffrom tensorflow import kerasimport matplotlib as mplfrom matplotlib import pyplot as plt%matplotlib inlineimport sklearnimport numpy as npimport pandas as pdimport osim...

#CALLBACK(回调函数)import tensorflow as tffrom tensorflow import kerasimport matplotlib as mplfrom matplotlib import pyplot as plt%matplotlib inlineimport sklearnimport numpy as npimport pandas as...

# 归一化：# １:把数据变成(０，１)或者（1,1）之间的小数。主要是为了数据处理方便提出来的，把数据映射到0～1范围之内处理，更加便捷快速。# ２:把有量纲表达式变成无量纲表达式，便于不同单位或量级的指标能够进行比较和加权。# 归一化是一种简化计算的方式，即将有量纲的表达式，经过变换，化为无量纲的表达式，成为纯量。归一化算法有：1、线性变换y=(x-MinValue)/(M...

import tensorflow as tffrom tensorflow import kerasimport matplotlib as mplfrom matplotlib import pyplot as plt%matplotlib inlineimport sklearnimport numpy as npimport pandas as pdimport osim...