tang0057的博客

转载 Tensorflow--Day 1

import tensorflow as tf# Model parameters W = tf.Variable([.3], dtype=tf.float32) b = tf.Variable([-.3], dtype=tf.float32) # Model input and output x = tf.placeholder(tf.float32) linear_model = W * x +

原创 Algorithms

Programming Thoughts：Reasoning thought; Recursive though; Greedy thought; Enumeration thought; Divid and Conquer; Backtracking; Dynamic Programming; Probability thoughtSorting：Internal sort：All data ha

原创 文章标题

（笔记总结需要！目前粗糙潦草，后期会不断修改。若对您阅读造成不便，望请见谅！！）8. PCAA dimensionality reduction algorithm that can be able to significantly speed up unsupervised feature learning algorithm. More importantly, understanding PC

原创 Autoencoders

（笔记总结需要！目前粗糙潦草，后期会不断修改。若对您阅读造成不便，望请见谅！！）7.Autoencoders：compress features, same function like PCAs.

原创 CNN

（笔记总结需要！目前粗糙潦草，后期会不断修改。若对您阅读造成不便，望请见谅！！）6. CNN1.initialize paremeters and load data.–>2. Bulid model.–>3. compute loss.–>4. Backpropagation.–>5. adjust parametersFully conntected network: all inputs in

原创 Perceptron

（笔记总结需要！目前粗糙潦草，后期会不断修改。若对您阅读造成不便，望请见谅！！）5. Perceptron

原创 SVM

（笔记总结需要！目前粗糙潦草，后期会不断修改。若对您阅读造成不便，望请见谅！！）4. SVMSVM的学习策略是间隔最大化。SVM的学习算法是求解凸优化规划（convex quadratic programming）的最优化算法。三种由简至繁的模型：Linear SVM in linearly separable case, linear SVM 和Nonlinear SVM。

原创 Boost

3. Boost AlgorithmsAdaBoost算法： 综合多个分类器，得到更精确结果。AdaBoost思想： 1、多轮训练，多个分类器 2、每轮训练增加错误样本的权值，降低正确分类样本权值。 3、降低错误率高的分类器权值，增加正确率高的分类器权值。

原创 KNN

（笔记总结需要！目前粗糙潦草，后期会不断修改。若对您造成不便，望请见谅！！）2. KNN参考链接：http://www.hankcs.com/ml/k-nearest-neighbor-method.html2.1 KNN算法给定一个训练数据集，对新的输入实例，在训练数据集中找到跟它最近的k个实例，根据这k个实例的类判断它自己的类（一般采用多数表决的方法）。KNN模型： 三要素：距离度量方法、K值

转载 Decision Tree

1. Decision TreeReference：http://www.hankcs.com/ml/decision-tree.html代码用ID3生成算法和C4.5生成算法实现，并用matplotlib可视化出来。分类决策树是对实例进行分类的树形结构。由结点和有向边组成。结点分为内部结点和叶节点。内部结点表示一个特征或属性，叶节点表示一个类。分类时，从根节点出发，当前结点设为根节点，且当前结点必

原创 Dynamic Programming--Algorithms

EditDistance(A[1 . . . n], B [1 . . . m]) D(i,0) ← i and D(0,j) ← j for all i,j for j from 1 to m:   for i from 1 to n:     insertion ← D (i , j − 1) + 1     delet

原创 Greatest common divisors & Fibonacci Numbers--Algorithms

Greatest Common Divisors: Naive Algorithm:      Function NaiveGCD(a,b) best ← 0 for d from 1 to a+b:   if d|a and d|b:      best ← d return best

原创 Graphs--Data Structure

Definition An (undirected) Graph is a collection V of vertices, and a collection E of edges each of which connects a pair of vertices. Vertices: A,B,C,D

原创 Greedy Algorithm--Algorithms

Largest Number:   Find max digit   Append it to the number   Remove it from the list of digits   Repeat while there are digits in the list  car fueling:   Refill at the

原创 AVL tree--Data Structure

Height Definition The height of a node is the maximum depth of its subtree. AVL Property AVL trees maintain the following propert

原创 Binary Search Tree--Data Structure

Hash table RangeSearch: impossible NearestNeighbors: impossible Insert:   O(1) Delete:  O(1) Array: RangeSearch:        O(n) NearestNeighbors: O(n) Insert:                     O(1) Delete

原创 Hash Tables: String Search--Data Structure

Naive Algorithm For each position i from 0 to |T| − |P|,check character-by-character whetherT[i..i + |P| − 1] = P or not. If yes, append i to the result.

原创 Hash Tables: Distributed Hash Tables--Data Structure

upload file instantly. Naive Comparison   Upload new file   Go through all stored files   Compare each stored file with new file byte-by-byte   If there's the same file, sto

原创 Hash Table: Hash Functions--Data Structure

Phone Book Design a data structure to store your contacts: names of people along with their phone numbers. The data structure should be able to do the following quickly:     Add and

原创 Basics of Hash Table--Data Structure

Intro: API: Python: Dict; JAVA: HashMap Applications: File Systems; Password Verification; Store Optimization IP Address:  Main Loop log - array of log lines (tim

原创 Disjoint Sets--Data Structure

Naive algorithm: A disjoint-set data structure supports the following operations:   MakeSet(x) creates a singleton set {x}   Find(x) returns ID of the set containing x:       if x and y lie in the

原创 Priority Queue--Data Structure

(一) Main operations of a queue:     PushBack(e); PopFront(). Priority queue is a generalization of a queue where each element is assigned a priority and elements come out in order by priority.

原创 Dynamic Array--Data Structure

Time complexity of Runtime: Get(i)                O(1) Set(i, val)          O(1) PushBack(val)  O(n) Remove(i)         O(n) Size()                O(1) Common Implementations: C++ : Vector

原创 DFS & BFS--Data Structure

Two ways of traversal : DFS, BFS Three methods to implement DFS: InOrderTraversal (tree)   if (tree == null) return;   InOrderTraversal (tree.left);   Print (tree.key);   InOrderTraversal (t

原创 Basic Tree--Data Structure

Pseudocode: Height (tree) if (tree == null) return 0; return 1 + Max ( Height ( tree.left ), Height ( tree.right )); Size (tree) if (tree == null) return 0; return 1 + Size ( tree.left) +Siz

原创 Stack & Queue--Data Structure

Stack： Queue： 在队列中，如果enqueue操作因为长度限制而从左向右压入数据，保留一个空间作为首尾的区分。

原创 Double Linked List--Data Structure

原创 Single Linked List--Data Structure

单链表形式如上图所示，内部包含“key”和“next pointer”指针，且指向一个方向。 API操作的时间复杂度：

转载 Bit Manipulation

常用位操作符有： For JAVA: 运算符优先级为从上到下递减，但>，>>>优先级相同。 操作符 功能 用法 ~ 位求反 ~expr 左移 expr1  >> 带符号右移，左边空出的位以原来最左边的0或者1填充 expr1 >> expr2 >>>

转载 JAVA "Public", "Protected", "Private" and "Friendly"的区别--笔记

1、public：public表明该数据成员、成员函数是对所有用户开放的，所有用户都可直接进行调用  2、protected：protected对于子类来说，就是public的，没有任何限制，而对于其他的外部class，protected就变成private。  3、private：private表示私有的意思。除class自己外，任何人都不可以直接使用，即便是子类都不可以使用。