Qian Hu

Keep curious to the world, and do cool things that matter.

  • 主页
  • 历史归档
  • 我的读书
  • 我的电影
  • 关于自己
所有文章 关于我

  • 主页
  • 历史归档
  • 我的读书
  • 我的电影
  • 关于自己

「论文学习」Cognitive Graph for Multi-Hop Reading Comprehension at Scale

2020-07-25

Basic Info

Link

Cognitive Graph for Multi-Hop Reading Comprehension at Scale

Author

  • Ming Ding
  • Chang Zhou
  • Qibin Chen
  • Hongxia Yang
  • Jie Tang
more >>
  • Homework
  • AI
  • Notes
  • Deep Learning
  • CogQA
  • Paper
  • Paper Notes

展开全文 >>

「论文学习」Semi-supervised Learning on Graphs with Generative Adversarial Nets

2020-07-25

Basic Info

Link

Semi-supervised Learning on Graphs with Generative Adversarial Nets

Author

  • Ming Ding
  • Jie Tang
  • Jie Zhang
more >>
  • Homework
  • AI
  • Notes
  • Deep Learning
  • Paper
  • GAN
  • Semi-supervised Learning
  • Paper Notes

展开全文 >>

「论文学习」DeepInf: Social Influence Prediction with Deep Learning

2020-07-18

Basic Info

Link

DeepInf: Social Influence Prediction with Deep Learning

Author

  • Jiezhong Qiu
  • Jian Tang
  • Hao Ma
  • Yuxiao Dong
  • Kuansan Wang
  • Jie Tang

Method(s)

Architecture

(a): Raw input which consists of a mini-batch of B instances; Each instance is a sub-network comprised n users who are sampled from the whole network. (b): An embedding layer which maps each user to a D-dimensional representation. (c): An Instance Normalization Layer which normalizes user’s embedding. (d): The formal input layer which concatenates together network embedding and other features. (e) A GCN and GAT Layer. (f) and (g) Compare model output and ground truth, we get the negative log likelihood loss. In this example, ego user v was finally activated (marked as black)

Reproduce

more >>
  • Homework
  • AI
  • Notes
  • Deep Learning
  • Paper
  • GCN
  • Social Influence
  • Social Network
  • Paper Notes

展开全文 >>

Kernel Functions for Support Vector Machines

2020-07-11

1. Background

Given input dimension d, input data x, y can be expressed as →x=(x1,...,xd), →y=(y1,...,yd), an SVM kernel function k(→x,→y) can be expressed as the dot product of the transformation of →x, →y by a transformation function γ, that is to say

k(→x,→y)=γ(→x)⋅γ(→y)

2. Problems

Derive the transformation function γ for the following SVM kernels, also compute the VC Dimension for the SVM model based on these kernels.

  • k(→x,→y)=(→x⋅→y)n
  • k(→x,→y)=(→x⋅→y+1)n
  • k(→x,→y)=e−σ(||→x||2+||→y||2)n

3. k(→x,→y)=(→x⋅→y)n

根据多项式展开公式

k(→x,→y)=(→x⋅→y)n=(x1y1+x2y2+⋯+xdyd)n=∑n!n1!n2!⋯nd!(x1y1)n1(x2y2)n2⋯(xdyd)nd

所以

more >>
  • Homework
  • Machine Learning
  • Kernel Function
  • SVM
  • Machine Learning

展开全文 >>

几个特殊的极限存在问题

2020-07-09

1. an=n∑i=11i=1+12+...+1n,数列{an}发散

an=n∑i=11i=1+12+...+1n=1+12+(13+14)+(15+16+17+18)+...>1+12+(14+14)+(18+18+18+18)+...+1n=1+12+12+...=1+m2​

其中m是12的个数,随着n的增大而增大

所以,

limn→∞an=∞

2. |an+p−an|≤pn,∀n,p∈N∗,数列{an}发散

举反例,如数列{1+12+...+1n}满足|an+p−an|≤pn,∀n,p∈N∗,但是其发散

3. |an+p−an|≤pn2,∀n,p∈N∗,数列{an}收敛

∀n,p∈N∗

|an+p−an|≤|an+p−an+p+1|+|an+p−1−an+p−2|+...+|an+1−an|≤1(n+p−1)2+...+1n2≤1(n+p−1)(n+p−2)+...+1n(n−1)=1n−1−1n+p−1<1n−1

more >>
  • Mathematics
  • Limit
  • Sequence Convergence
  • Calculus
  • Mathematical Analysis
  • Mathematical Analysis

展开全文 >>

VC Dimension of Neural Networks

2020-06-19

1. Problems

We take 0/1 classification problem for data with d dimensional features as an example. A neural network with one hidden layer can be written as

o=wT2σ(W1v+b1)+b2

where v is the d dimensional input feature of the data, while W1, w2, b1, b2 are the parameters of the neural network model. W1 is a n×d matrix, w2 is a n dimensional vector, and b1 is a n dimensional bias vector while b2 is the bias. When o>0 we classify the datum as label 1, while when o≤0 we classify it as label -1. This forms a neural network, or multi-layer-perceptron, with one hidden layer containing n neurons. In this problem, we focus on the pre-training case with frozen parameters that W1 and b1 must be decided when seeing all v without labels (l1,...li), while w2 and b2 can be decided after seeing all labels of examples (l1,...li)

1.1 Problem1

Given n,d, calculate the VC dimension of the neural network for the linear activation case, i.e. σ(x)=x. Prove your result.

1.2 Problem2

Given n,d, calculate the VC dimension of the neural network for the ReLU activation case, i.e. σ(x)=max(0,x). Prove your result.

1.3 Hints

  • Recall the definition of VC dimension.
  • Consider n>d and n≤d.
  • For problem 2, Start from d=1
more >>
  • Homework
  • Machine Learning
  • VC Dimension
  • Machine Learning

展开全文 >>

「论文笔记」ImageNet Classification with Deep Convolutional Neural Networks

2020-03-28

Basic Info

Link: ImageNet Classification with Deep Convolutional Neural Networks - NIPS

Author:

  • Alex Krizhevsky, University of Toronto
  • Ilya Sutskever, University of Toronto
  • Geoffrey E. Hinton, University of Toronto

Summary

Research Objective

作者的研究目标。

Problem Statement

问题陈述,要解决什么问题?

Method(s)

more >>
  • AI
  • Notes
  • Deep Learning
  • Paper
  • Computer Vision
  • Object Detection
  • Comments
  • Paper Notes

展开全文 >>

IOS学习笔记

2019-11-06

对C的扩展

#import

可保证头文件只被包含一次,无论此命令在那个文件中出现了多少次

NSLog 和@字符串

NSLog类似于C语言的printf(),添加了例如时间戳、日期戳和自动附加换行符等特性

NS前缀是来自Cocoa的函数,@符号表示引用的字符串应作为Cocoa的NSString元素来处理

NSArray提供数组,NSDateFormatter帮助用不同方式来格式化日期,NSThread提供多线程编程工具,NSSpeechSynthesizer使听到语音

使用NSLog()输出任意对象的值时,都会使用%@格式说明,使用这个说明符时,对象通过一个名为description的方法提供自己的NSLog()形式

Boolean类型

相比于C语言的bool类型,OC提供BOOL类型,具有YES(1)值和NO(0)值,编译器将BOOL认作8位二进制数

more >>

展开全文 >>

「论文笔记」Rich feature hierarchies for accurate object detection and semantic segmentation

2019-08-01

Basic Info

Link: arXiv:1311.2524

Author: Ross Girshick, Jeff Donahue, Trevor Darrell, Jitendra Malik

Summary

Research Objective

作者的研究目标。

Problem Statement

问题陈述,要解决什么问题?

Method(s)

解决问题的方法/算法是什么?

more >>
  • AI
  • Notes
  • Deep Learning
  • Paper
  • Computer Vision
  • Object Detection
  • Comments
  • Paper Notes

展开全文 >>

「西瓜书学习」Ch03.线性模型

2019-05-30

1. 基本形式

给定由d个属性描述的示例x=(x1;x2;…;xd),其中xi是x在第i个属性上的取值,线性模型试图学得一个通过属性的线性组合来进行预测的函数,即

f(x)=w1x1+w2x2+…+wdxd+b

一般用向量形式写成

f(x)=wTx+b

其中w=(w1;w2;…;wd)。w和b学得之后,模型就得以确定

2. 线性回归

给定数据集D=(x1,y1),(x2,y2),…,(xm,ym),其中xi=(xi1,xi2,…,xid),yi∈R,"线性回归"(linear regression)试图学得一个线性模型以尽可能准确地预测实值输出标记。

我们先考虑一种最简单的情形:输入属性的数目只有一个。为便于讨论,此时我们忽略关于属性的下标,即D={(xi,yi)}mi=1,其中xi∈R,对于离散属性,若属性值之间存在"序"(order)的关系,可通过连续化将其转化为连续值;若属性值间不存在序关系,假定有k个属性值,则通常转化为k维向量。

线性回归试图学得

more >>
  • Machine Learning
  • AI
  • Notes

展开全文 >>

&laquo; Prev12345…7Next &raquo;
© 2024 Qian Hu
Hexo Theme Yilia by Litten
  • 所有文章
  • 关于我

tag:

  • Linux
  • Tensorflow Serving
  • Nvidia
  • CUDA
  • CUDNN
  • Inference Speed
  • Hexo
  • LaTex
  • Homework
  • Machine Learning
  • Kernel Function
  • SVM
  • Algorithm
  • LeetCode
  • Interview
  • Hash
  • Double Linked List
  • BinarySearch
  • TopologicalSort
  • List
  • Shell
  • Python
  • File Processing
  • MySQL
  • Database
  • PAT
  • Design Patterns
  • Magic Function
  • Decorater
  • Translations
  • StackOverflow
  • VC Dimension
  • Datawhale
  • AI
  • Data Mining
  • LogisticRegression
  • DecisionTree
  • Practice
  • RandomForest
  • GBDT
  • XGBoost
  • LightGBM
  • Model Evaluation
  • Grid Search
  • Software Architecture
  • GeekTime
  • Study Record
  • Wechat
  • Notes
  • Deep Learning
  • CogQA
  • Paper
  • GCN
  • Social Influence
  • Social Network
  • GAN
  • Semi-supervised Learning
  • Computer Vision
  • Object Detection
  • Comments
  • Movie
  • Mathematics
  • Limit
  • Sequence Convergence
  • Calculus
  • Mathematical Analysis
  • Evernote
  • tools
  • system code
  • DataStructure
  • BinarryTree
  • Programing
  • Programing Language
  • Go
  • Math
  • Dynamic Programming
  • Encrypt
  • Operating System
  • Network
  • I/O

    缺失模块。
    1、请确保node版本大于6.2
    2、在博客根目录(注意不是yilia根目录)执行以下命令:
    npm i hexo-generator-json-content --save

    3、在根目录_config.yml里添加配置:

      jsonContent:
        meta: false
        pages: false
        posts:
          title: true
          date: true
          path: true
          text: false
          raw: false
          content: false
          slug: false
          updated: false
          comments: false
          link: false
          permalink: false
          excerpt: false
          categories: false
          tags: true
    

  • I/O多路复用select/poll/epoll

    2024-08-19

    #Operating System#Network#I/O

  • 对称加密与非对称加密

    2024-08-19

    #Algorithm#Encrypt

  • HTTPS原理和应用

    2024-08-19

  • LeetCode - 70.Climbing Stair

    2024-08-17

    #Algorithm#LeetCode#Interview#Math#Dynamic Programming

  • Redis的基本数据类型和数据结构

    2024-08-14

  • Go并发编程最佳实践习题题解

    2024-07-28

    #Homework#Study Record#Programing Language#Go

  • Go神奇的内置数据结构习题题解

    2024-07-25

    #Homework#Study Record#Programing Language#Go

  • Go语法背后的秘密作业题解

    2024-07-23

    #Homework#Study Record#Programing Language#Go

  • Go runtime相关函数源码阅读

    2024-07-22

    #Homework#Study Record#Programing Language#Go

  • 「架构实战营」作业二 微信朋友圈高性能复杂度分析

    2021-04-18

    #Homework#Software Architecture#GeekTime#Study Record#Wechat

  • 「架构实战营」作业一 微信业务架构图 & 毕设版学生管理系统架构设计

    2021-04-07

    #Homework#Software Architecture#GeekTime#Study Record#Wechat

  • 腾讯极客挑战赛第三期-码上种数Writeup

    2021-03-13

    #Programing

  • LeetCode反转链表相关题整理

    2021-02-07

    #Algorithm#LeetCode#Interview#List

  • Flask框架的基本原理

    2021-02-05

  • LeetCode - 146.LRU Cache

    2021-02-05

    #Algorithm#LeetCode#Interview#Hash#Double Linked List

  • MySQL下的事务隔离级别及可能出现的问题

    2021-01-30

    #MySQL#Database

  • Python下实现单例模式

    2021-01-26

    #Python#Design Patterns#Magic Function#Decorater

  • Linux下的文件批量重命名

    2021-01-22

    #Linux#Shell#Python#File Processing

  • CUDA升级,从9.2到11.0

    2021-01-20

    #Linux#Tensorflow Serving#Nvidia#CUDA#CUDNN#Inference Speed

  • 印象笔记存储导致的tensorflow serving的坑 Error parsing text-format tensorflow.serving.ModelServerConfig

    2020-09-26

    #Linux#Tensorflow Serving#Evernote#tools#system code

  • 「论文学习」Cognitive Graph for Multi-Hop Reading Comprehension at Scale

    2020-07-25

    #Homework#AI#Notes#Deep Learning#CogQA#Paper

  • 「论文学习」Semi-supervised Learning on Graphs with Generative Adversarial Nets

    2020-07-25

    #Homework#AI#Notes#Deep Learning#Paper#GAN#Semi-supervised Learning

  • 「论文学习」DeepInf: Social Influence Prediction with Deep Learning

    2020-07-18

    #Homework#AI#Notes#Deep Learning#Paper#GCN#Social Influence#Social Network

  • Kernel Functions for Support Vector Machines

    2020-07-11

    #Homework#Machine Learning#Kernel Function#SVM

  • 几个特殊的极限存在问题

    2020-07-09

    #Mathematics#Limit#Sequence Convergence#Calculus#Mathematical Analysis

  • VC Dimension of Neural Networks

    2020-06-19

    #Homework#Machine Learning#VC Dimension

  • 「论文笔记」ImageNet Classification with Deep Convolutional Neural Networks

    2020-03-28

    #AI#Notes#Deep Learning#Paper#Computer Vision#Object Detection#Comments

  • IOS学习笔记

    2019-11-06

  • 「论文笔记」Rich feature hierarchies for accurate object detection and semantic segmentation

    2019-08-01

    #AI#Notes#Deep Learning#Paper#Computer Vision#Object Detection#Comments

  • 「西瓜书学习」Ch03.线性模型

    2019-05-30

    #Machine Learning#AI#Notes

  • 「MySQL基础」Day6 复杂项目

    2019-03-04

    #MySQL#Datawhale

  • 「MySQL基础」Day5 MySQL 实战

    2019-03-03

    #MySQL#Datawhale

  • 「MySQL基础」Day4 表联结

    2019-03-02

    #MySQL#Datawhale

  • 「MySQL基础」Day3 表操作

    2019-03-01

    #MySQL#Datawhale

  • 「MySQL基础」Day2 查询语句

    2019-02-27

    #MySQL#Datawhale

  • 「MySQL基础」Day1 软件安装与数据库基础

    2019-02-25

    #MySQL#Datawhale

  • 「统计学学习」Day4 回归分析与卡方分布

    2019-01-09

  • 「统计学学习」Day3 假设检验

    2019-01-07

  • 「统计学学习」Day2 中心极限定理与置信区间

    2019-01-05

  • 「统计学学习」Day1 基本知识与大数定律

    2019-01-04

  • Hexo渲染LaTex数学公式的方法

    2018-12-23

    #Hexo#LaTex

  • 「西瓜书学习」Ch02.模型评估与选择

    2018-12-23

    #Machine Learning#AI#Notes

  • 「一周算法实践」Task3 模型调优

    2018-12-15

    #AI#Data Mining#Practice#Model Evaluation#Grid Search

  • 「一周算法实践」Task2 模型评估

    2018-12-13

    #AI#Data Mining#Practice#Model Evaluation

  • LeetCode 207. Course Schedule

    2018-12-12

    #Algorithm#LeetCode#BinarySearch#TopologicalSort

  • 「一周算法实践」Task1.2 模型构建之集成模型

    2018-12-09

    #AI#Data Mining#Practice#RandomForest#GBDT#XGBoost#LightGBM

  • 「一周算法实践」Task1.1 模型构建

    2018-12-08

    #SVM#AI#Data Mining#LogisticRegression#DecisionTree#Practice

  • LeetCode 33.Search in Rotated Sorted

    2018-12-08

    #Algorithm#LeetCode#BinarySearch

  • 真二叉重构(Proper Rebuild)

    2018-12-05

    #Algorithm#DataStructure#BinarryTree

  • MySQL一些基础知识及存储引擎比较

    2018-01-09

    #MySQL#Database

  • Python中的yield关键字

    2017-12-30

    #Python#Translations#StackOverflow

  • 不仅仅是寻梦,也蕴含着亲情和温情的故事 -- 《寻梦环游记》影评

    2017-12-29

    #Comments#Movie

  • PAT-甲级 1002. A+B for Polynomials

    2017-12-28

    #Algorithm#PAT

  • PAT-甲级 1001.A+B Format

    2017-12-28

    #Algorithm#PAT

  • 在已有目录上添加Git

    2017-12-28

  • MySQL无法获取最新数据的解决方法

    2017-09-30

    #MySQL#Database

  • 《Effective Python》读书笔记

    2017-06-20

  • jQuery DataTables的增删改方案

    2017-06-13

  • 将linux下的rm改造成mv

    2017-05-26

  • ubuntu 16.04下搜狗输入法崩溃的解决方案

    2017-03-24

  • Python项目环境的搭建

    2017-03-23

  • Preface

    2017-03-22

Python后端开发,业余前端,自学CV中