机器学习
齐豪
...
展开
-
lstm
http://colah.github.io/posts/2015-08-Understanding-LSTMs/原创 2018-03-15 20:34:26 · 223 阅读 · 0 评论 -
机器学习资源
coursera的机器学习公开课 https://www.coursera.org/learn/machine-learning neural-networks-and-deep-learning https://hit-scir.gitbooks.io/neural-networks-and-deep-learning-zh_cn/content/ Udacity深度学习免费公原创 2017-08-21 17:24:30 · 724 阅读 · 0 评论 -
用神经网络分类数字代码
代码#!/usr/bin/python # -*- coding: UTF-8 -*- import random import cPickle import gzip import numpy as np class Network(object): def __init__(self, sizes): self.num_layers = len(sizes)原创 2017-08-21 17:27:29 · 1159 阅读 · 0 评论 -
神经网络反向传播算法
四个基本等式证明 算法代码 def backprop(self, x, y): """Return a tuple "(nabla_b, nabla_w)" representing the gradient for the cost function C_x. "nabla_b" and "nabla_w" are layer-by-laye原创 2017-08-21 17:25:58 · 333 阅读 · 0 评论 -
Softmax和交叉熵
Softmaxsoftmax(x)i=exp(xi)∑jexp(xj)\text{softmax}(x)_i = \frac{\exp(x_i)}{\sum_j \exp(x_j)}aLj=ezLj∑kezLk\begin{eqnarray} a^L_j = \frac{e^{z^L_j}}{\sum_k e^{z^L_k}}\end{eqnarray}∑jaLj=∑jezLj∑kezLk=1\be原创 2017-08-21 17:23:04 · 955 阅读 · 0 评论 -
Detection ouput
Detection ouput SSD 输入输出 inputs[0]: location map, {N, boxes * 4} inputs[1]: confidence map, ssd: {N, classes, boxes}, yolov3: {N, boxes, classes} inputs[2]: priorbox prior boxes, dims = 4 {1, 2,...原创 2019-05-20 17:53:49 · 248 阅读 · 0 评论