普林斯顿大学算法第一部分学习总结(Week1-Percolati

2017-06-23  本文已影响0人  自度君

Algorithms Part1课程第一周的Programming Assignment是Percolation问题,该问题是Union-Find算法的一个应用实例。作业要求见以下链接:http://coursera.cs.princeton.edu/algs4/assignments/percolation.html,要求用Java实现算法。此外该课程使用了作者提供的API用以简化编程,如基本输入、输出、随机数生成、数据分析等库函数,在上述链接中可直接下载。

模型描述:

Percolation即渗透过程,其模型如下:一个方形“水槽”(system)由N*N个方格(site)组成,每个方格有开启(open)或关闭(blocked)两个状态,相邻(即上下左右)的开启格子能够构成一条同路。如果一个格子处于开启状态,并且与顶端的一行能够通过开启的格子实现连通(connected)通路,我们说这个格子处于充满水(full)的状态;如果一个“水槽”(system)的顶端格子与底部格子通过开启的格子实现了连通,我们称这个“水槽”是可以渗透(percolates)的。如下图所示:

image.png

第一章的作业需要制作两个类。第一个Percolation包含五个方法,主要是用来打开渗漏的格子。第二个类PercolationStats 同样有五个方法,用来找出N*N格子中,能够产生渗漏的阈值概率的95%置信区间。先说一下大概的想法和原理。官网上提供有这道题目的思路,就是在创建格子的过程中首先要在顶层和底层各加一层,这样的话在测试是否渗漏时直接测试(0,1)和(N+1,1)就好了首先来说下第一个类:官网的API已经写好:

public class Percolation {  
   public Percolation(int N)               // create N-by-N grid, with all sites blocked  
   public void open(int i, int j)          // open site (row i, column j) if it is not open already  
   public boolean isOpen(int i, int j)     // is site (row i, column j) open?  
   public boolean isFull(int i, int j)     // is site (row i, column j) full?  
   public boolean percolates()             // does the system percolate?  
  
   public static void main(String[] args   // test client (optional)  
}  
  1. 构造函数 Percolation
    该构造方法对模型初始化,我们创建一个NN大小的一维布尔数组作为“水槽”,数组值记录每个格子的开闭状态,初始状态为关闭。此外,我们需要另一个(NN+2)大小的一维数组存储连通关系,根据Union-Find算法的学习,我们创建类型为WeightedQuickUnionUF的数组。大小之所以为(N*N+2),是因为在检查是否渗透的时候,可以通过在顶部与底部设置两个虚拟点,其与顶端或底部的一排是相连通的,这样检查是否渗透的时候,无需遍历顶部或底部一排的每一个点,只需要检查两个虚拟点是否连通即可,如下图所示:
    image.png
  2. 打开格子方法open
    检查坐标为(i, j)的点是否开启,如果没有开启,则将其开启。注意i,j的取值范围为1~N。此外,开启格子后,需要检查其上下左右的格子是否同样已经打开,如果已经打开,则通过WeightedQuickUnionUF的union()方法将两个格子连通起来;另外根据作业要求,需要首先检查是否越界,抛出java.lang.IndexOutOfBoundsException()错误;
  3. isOpen(int i, int j):只需要返回该点的开闭状态值;
  4. isFull(int i, int j):检查该点是否已经打开并且已经与顶部的虚拟点(相当于顶部的一排)连通,用WeightedQuickUnionUF的connected()方法检测;
  5. percolates():检查顶部的虚拟点与底部的虚拟点是否连通,用WeightedQuickUnionUF的connected()方法检测;

注意:

image.png
package percolation; /**
 * @author rxy8023
 * @version 1.0    2017/6/17 22:00
 */

import edu.princeton.cs.algs4.In;
import edu.princeton.cs.algs4.StdOut;
import edu.princeton.cs.algs4.WeightedQuickUnionUF;

public class Percolation {

    // number of grid should be n +1,because row and column indices are between 1 and n
    private int n;
    // union-find data structure
    private WeightedQuickUnionUF uf;
    // mark open site
    // 0 for Blocked site, 1 for Open site, 2 for Open site connected to the bottom
    private byte[][] open;
    // number of open sites
    private int num;

    /**
     * create n-by-n grid, with all sites blocked.
     *
     * @param n number of dimension
     */
    public Percolation(int n) {
        if (n <= 0) throw new IllegalArgumentException("Invalid input : n must > 0 !");
        this.n = n + 1;
        uf = new WeightedQuickUnionUF(this.n * this.n);
        open = new byte[this.n][this.n];
    }

    public static void main(String[] args) {
        In in = new In(args[0]);
        int n = in.readInt();
        Percolation percolation = new Percolation(n);
        boolean isPercolated = false;
        int count = 0;
        while (!in.isEmpty()) {
            int row = in.readInt();
            int col = in.readInt();
            if (!percolation.isOpen(row, col)) {
                count++;
            }
            percolation.open(row, col);
            isPercolated = percolation.percolates();
            if (isPercolated) {
                break;
            }
        }
        StdOut.println(count + " open sites");
        if (isPercolated) {
            StdOut.println("percolates");
        } else {
            StdOut.println("does not percolate");
        }

    }

    /**
     * Validate the row and col indices.
     *
     * @param row row index
     * @param col col index
     */
    private void validate(int row, int col) {
        if (row <= 0 || row >= n) {
            throw new IndexOutOfBoundsException("Invalid input : row index out of bounds !");
        }
        if (col <= 0 || row >= n) {
            throw new IndexOutOfBoundsException("Invalid input : col index out of bounds !");
        }
    }

    /**
     * open site (row, col) if it is not open already
     *
     * @param row the index of row
     * @param col the index of column
     */
    public void open(int row, int col) {
        validate(row, col);
        // is open already
        if (isOpen(row, col)) return;
        // make this site open
        open[row][col] = 1;
        num++;
        // we make 0 represent the virtual-top, 1 represent the virtual-bottom.
        // is the bottom row
        if (row == n -1) open[row][col] = 2;
        // is the top row
        if (row == 1) {
            uf.union(0, row * n + col);
            // 1-by-1 grid corner case
            if (open[row][col] == 2) open[0][0] = 2;
        }
        // above site is open
        if (row - 1 > 0 && isOpen(row - 1, col)) {
            update(row - 1, col, row, col);
        }
        // below site is open
        if (row + 1 < n && isOpen(row + 1, col)) {
            update(row + 1, col, row, col);
        }
        // left site is open
        if (col - 1 > 0 && isOpen(row, col - 1)) {
            update(row, col - 1, row, col);
        }
        // right site is open
        if (col + 1 < n && isOpen(row, col + 1)) {
            update(row, col + 1, row, col);
        }
    }

    /**
     * update components: connect the opened site to all of its adjacent open sites
     *
     * @param i1 adjacent open site row index
     * @param j1 adjacent open site col index
     * @param i2 the opened site row index
     * @param j2 the opened site col index
     */
    private void update(int i1, int j1, int i2, int j2) {
        int p = uf.find(i1 * n + j1);
        int q = uf.find(i2 * n + j2);
        uf.union(i1 * n + j1, i2 * n + j2);
        // if one of them is connected to bottom, then the updated component is connected to bottom too.
        if (open[p / n][p % n] == 2 || open[q / n][q % n] == 2) {
            int t = uf.find(i2 * n + j2);
            open[t / n][t % n] = 2;
        }
    }

    /**
     * is site (row, col) open?
     *
     * @param row row index
     * @param col col index
     * @return
     */
    public boolean isOpen(int row, int col) {
        validate(row, col);
        return open[row][col] > 0;
    }

    /**
     * is site (row, col) full?
     *
     * @param row row index
     * @param col col index
     * @return
     */
    public boolean isFull(int row, int col) {
        validate(row, col);
        return open[row][col] > 0 && uf.connected(0, row * n + col);
    }

    /**
     * number of open sites
     *
     * @return
     */
    public int numberOfOpenSites() {
        return num;
    }

    /**
     * does the system percolate?
     *
     * @return
     */
    public boolean percolates() {
        int root = uf.find(0);
        return open[root / n][root % n] == 2;
    }


}

蒙特卡洛仿真分析(PercolationStats类)

用蒙特卡洛(Monte Carlo)方法进行仿真分析,确定渗透阈值p,具体方法是:
初始化将所有N*N个格子置于关闭状态;
每次随机选择一个格子进行开启操作,直到系统达到渗透状态;
统计此时开启的格子数,与全部格子数相比计算一个p值的样本;
独立重复T次上述实验,最后计算出T个p值的样本,记为x1,x2……xT。

进行以下统计操作。
按照下式计算均值与方差:



假设T值足够大(至少30次),则可以根据下式计算置信区间在95%的的渗透阈值范围:


package percolation;

import edu.princeton.cs.algs4.StdOut;
import edu.princeton.cs.algs4.StdRandom;
import edu.princeton.cs.algs4.StdStats;

/**
 * @author rxy8023
 * @version 1.0    2017/6/17 22:00
 */
public class PercolationStats {
    // sample mean of percolation threshold
    private double mean;
    // sample standard deviation of percolation threshold
    private double stddev;
    // low  endpoint of 95% confidence interval
    private double confidenceLo;
    // high endpoint of 95% confidence interval
    private double confidenceHi;
    // est[i] = estimate of percolation threshold in perc[i]
    private double[] est;

    /**
     * perform trials independent experiments on an n-by-n grid
     *
     * @param n
     * @param trials
     */
    public PercolationStats(int n, int trials) {
        if (n <= 0 || trials <= 0) throw new IllegalArgumentException("Invalid input : n or trials musu > 0 !");
        est = new double[trials];
        for (int k = 0; k < trials; k++) {
            Percolation perc = new Percolation(n);
            double count = 0;
            while (!perc.percolates()) {
                int i = StdRandom.uniform(1, n + 1);
                int j = StdRandom.uniform(1, n + 1);
                if (perc.isOpen(i, j)) continue;
                perc.open(i, j);
                count++;
            }
            est[k] = count / (n * n);
        }
        mean = StdStats.mean(est);
        stddev = StdStats.stddev(est);
        confidenceLo = mean - (1.96 * stddev) / Math.sqrt(trials);
        confidenceHi = mean + (1.96 * stddev) / Math.sqrt(trials);
    }

    public static void main(String[] args) {
        int n = Integer.parseInt(args[0]);
        int t = Integer.parseInt(args[1]);
        PercolationStats stats = new PercolationStats(n, t);
        StdOut.println("mean                    = " + stats.mean());
        StdOut.println("stddev                  = " + stats.stddev());
        StdOut.println("95% confidence interval = " + stats.confidenceLo()
                + ", " + stats.confidenceHi());
    }

    /**
     * Get sample mean of percolation threshold
     *
     * @return
     */
    public double mean() {
        return mean;
    }

    /**
     * Get sample standard deviation of percolation threshold
     *
     * @return
     */
    public double stddev() {
        return stddev;
    }


    /**
     * Get low  endpoint of 95% confidence interval
     *
     * @return
     */
    public double confidenceLo() {
        return confidenceLo;
    }


    /**
     * Get high endpoint of 95% confidence interval
     *
     * @return
     */
    public double confidenceHi() {
        return confidenceHi;
    }
}

总结:

如何根据API编译JAVA程序:

1.把API所需要做的内容全都用文字详细写出来,每个步骤该做什么;
2.可以写出伪代码,类似于算法课上所要求的,详尽写出;
3.选好数据结构,简单变成不需要;
4.Coding

上一篇下一篇

猜你喜欢

热点阅读