TensorFlow 的计算需要事先定义一个 computation graph（计算图），该图是一个抽象的结构，只有在评估（evaluate）时，才有数值解，这点和 numpy 不同。这张图由一组节点构成。
>> a = tf.ones((2, 2))
>> tf.reduce_sum(a, reduction_indices=1)
<tf.Tensor 'S
转载
2017-03-15 12:10:00
103阅读
2评论
Overflow and UnderflowThe fundamental difficulty in performing continuous math on a digital computer is
原创
2021-07-09 14:20:44
98阅读
COMPUTER OR GANIZATION AND ARCHITECTURE DESIGNING FOR PERFORMANCE NINTH EDITION Although the performance of mainframe general-purpose computers contin
转载
2017-01-05 20:34:00
75阅读
2评论
Overflow and UnderflowThe fundamental difficulty in performing continuous
原创
2022-01-19 09:30:31
157阅读
#include <iostream>#include <string>#include <algorithm>using namespace std;template<int N>class aTMP{public: enum { ret = N * aTMP<N-1>::ret };...
原创
2022-12-01 17:00:23
62阅读
# 实现"Computation Thread Pool Size"
作为一名经验丰富的开发者，你可能经常需要处理大量的计算任务。为了提高效率，你需要将这些任务分配给多个线程进行并发处理。但是线程的数量也不能无限增长，因为过多的线程会导致系统资源的浪费和竞争。所以，为了避免这种情况，我们需要使用线程池来管理线程的数量。
## 线程池的概念
线程池是一种管理多个线程的技术，它可以预先创建一定数量的
原创
2024-01-15 21:40:12
41阅读
介绍一篇论文《Nectar:Automatic Management of Data and Computation in Datacenters》
转载
2021-07-13 16:13:49
56阅读
超计算（Hyper computation）模型 作者：Xyan Xcllet链接：https://www.zhihu.com/question/21579465/answer/106995708来源：知乎著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。 超计算，是一个研究比图灵
转载
2017-11-07 18:29:00
703阅读
Abstract. We present a general method to compile any cryptographicalgorithm into one which resists side channel attacks of the only computation leaks information variety for an unbounded number of ex
原创
2024-06-04 00:21:39
82阅读
Preface First, let’s recap, there are 2 kinds of objects in .NET: value types and reference types that are created on the stack and in the heap (manag
转载
2020-11-04 00:16:00
125阅读
2评论
<br /> <br />The purpose of this Note is to document default CSS misscount timeout c
原创
2022-09-02 08:57:58
125阅读
Graph There are two standard ways to represent a graph G=(V,E)G=(V,E), where VV is a set of vertices and EE is a set of edges; Adjacency list represen
转载
2019-04-30 16:26:00
339阅读
2评论
The context managers torch.no_grad(), torch.enable_grad(), and torch.set_grad_enabled() are helpful for locally disabling and enabling gradient computation. See Locally disabling gradient computation ...
原创
2021-08-12 22:30:33
151阅读
转载
2017-10-30 20:56:00
43阅读
2评论
uted File System (HDFS) is a distributed file system designed to
转载
2017-10-19 10:45:00
67阅读
2评论
This lesson will teach you about AsyncSubject, another type of Subject with some replaying logic inside. We will also look at some use cases for this
转载
2016-10-19 15:30:00
50阅读
2评论
Abstract. We study the following two related questions:
{ What are the minimal computational resources required for general
secure multiparty computation in the presence of an honest majority?
{ What
原创
2024-03-27 11:05:34
49阅读
Graph( graph .cpp/c/pas)Description小 Y 又开始了一段旅途。这次，他要经过一个图，从1号点到达n号点，每个点设有休息站。小 Y 计划用最多k天走完全程，除第k天外，每一天小 Y 都必须在休息站过夜。所以，一段路必须在同一天走完。小 Y 的体力有限，他希望走的路程最
转载
2016-08-22 10:38:00
168阅读
2评论
GraphThe gt-graph package defines the concept of a graph (or network) made up of GeoTools Features.gt-graph包定义了由GeoTools要素构成的图（或网络）的概念。Maven:<dependency>
<groupId>org.geotools&
转载
2024-05-09 23:34:52
192阅读
maxpooling 的 max 函数关于某变量的偏导也是分段的，关于它就是 1，不关于它就是 0；
BP 是反向传播求关于参数的偏导，SGD 则是梯度更新，是优化算法；
1. 一个实例
relu = tf.nn.relu(tf.matmul(x, W) + b)
C = [...]
[db, dW, dx] = tf.gradient(C, [b, w, x])
转载
2017-03-16 11:29:00
421阅读
2评论
