Description
Question 1: Warm Up Math Questions [30] a) Take the derivative with respect to x : f(x) = ln(4 + sin2 x) + e 3x cos x b) Two 6-sided dice are rolled. Find the probability that the sum is less than 11. c) Find the value of: limx→0 (e x − 1)(sin x) x ln(x + 1) Question 2: Hierarchical Clustering [20] The following table shows the distance between some cities within the state of Wisconsin. The distances are given in miles. Distance (Miles) City MA MI AP CH GC Madison (MA) 0 78.7 105.8 216.1 223.8 Milwaukee (MI) 78.7 0 107.2 283.1 291.5 Appleton (AP) 105.8 107.2 0 222.4 239.3 Chetek (CH) 216.1 283.1 222.4 0 46.0 Glenwood City (GC) 223.8 291.5 239.3 46.0 0 Use Hierarchical Clustering with single-linkage to cluster the cities by hand, until all cities are in the same cluster. 1. For each iteration, show the cluster membership. 2. For each iteration, show all pairwise distances between clusters. Question 3 Programming: K-Means [50] In this question, you will be building a k-means clustering algorithm with Euclidean distance. We are providing a skeleton code ”KMeans.zip” for you at the course Web page. Please download, unzip this file, and open the file KMeans.java, you will see: public c l a s s KMeans { public KMeansResult c l u s t e r ( double [ ] [ ] c e n t r oi d s , double [ ] [ ] i n s t a n c e s , double t h r e s h ol d ) { /∗ . . . YOUR CODE GOES HERE . . . ∗/ } } CS 540-1 Fall 2014 Your contribution is the implementation of this cluster method. It accepts an array of initial centroids, an array of instances, and a threshold for stopping the iterations of the algorithm. It must return an object of type: public c l a s s KMeansResult { double [ ] [ ] c e n t r o i d s ; double [ ] d i s t o r t i o n I t e r a t i o n s ; int [ ] cl u s t e rA s si g nm e n t ; } which is defined in KMeansResult.java. We will use this object to verify the correctness of your clustering algorithm. The centroids is an array of the final coordinate of your centroids. The distortionIterations will return an array of each successive distortion your algorithm records on each iteration. This allows us to know both how many iterations your algorithm required and how well it clustered the data set before converging. When the difference between successive distortions, recorded on each iteration, is less than the threshold argument, your program should stop iterating and return an instance of KMeansResult that provides us with the appropriate arrays. More about KMeans.cluster() Since KMeans.cluster() is the core part of this question, we will give a more detailed explanation of its prototype. Parameters You are required to pass three arguments to the method cluster . Their detailed meanings are given as followed: • double[][] centroids A two-dimensional array of the initial position of the centroids. Each row corresponds to a centroid and each column to a feature dimension. • double[][] instances The data set you will be clustering. Each row corresponds to an instance, and this array will have the same “width” as the centroids array. • double threshold The threshold that determined when to stop iterations. More specially, if at iteration i the relative change in distortion between successive iterations is less than the threshold, i.e. distortion(i) − distortion(i − 1) distortion(i − 1) < threshold, then your algorithm should terminate and return its results. Returns You are required to return a variable with the type KMeansResult. The detailed meanings of its each field are given as followed: • double[][] centroids The position of the final centroids after your clustering is complete. CS 540-1 Fall 2014 • double[] distortionIterations An array of the distortions your algorithm records on successive iterations • int[] clusterAssignment An array of the index of the centroid assigned to each row of the instances method argument. That is, the integers in this array will point to a row in the returned double[][] centroids array. So, for example, if we wished to know the coordinate, in the feature space, of the centroid assigned to the first instance, we would call centroids[clusterAssignment[0]] . Remember that all Java arrays are 0-based. Implementation details In each iteration, you should first reallocate the clusters for each instance, then update the coordinates of centroids by averaging all instances in their corresponding clusters. It is possible that, in some iteration after reallocation, there is a centroid c which does not match to any instances. In this case, we call the centroid of this cluster orphaned centroid. This is a problem, because you would have no instances assigned to the centroid to average over to determine the centroid’s new location. To solve this, we specify you to implement the following behaviors in this scenario: 1. Search among all the instances for the instance x whose distance is farthest from its assigned centroid. 2. Choose x’s position as the position of c, the orphaned centroid. 3. Reallocate again the cluster assignments for all x. 4. Check if there is still a orphaned centroid. If so repeat step 1 to 4 until all orphaned centroids have been removed. Once you have removed all possible orphaned centroids, you should update the centroids’ coordinates by averaging over all the instances assigned to it. After that, you are also required to calculate the distortion of an iteration, and store it into the array distortionIterations. Notice that we don’t know the number of iterations before hand, so you may need to use a variable-length array like Vector or ArrayList first, and write the result into distortionIterations after all iterations. Moreover, it is also possible that the minimum or maximum distance may correspond to multiple candidates. For example, there may be more than one cluster centroids have the same minimum distance to an instance, or several instances have the same maximum distance to their corresponding centroids. In this case, always choose the one with smaller id, either instance id, cluster id, or so. Finally, your program should be able to deal with different number of centroids, instances, as well as the feature dimensions, and work correctly no matter what the size of the double[][] centroids and double[][] instances arrays. So do not hard-code any array lengths. The .length member allows you to determine the length of the array. You can determine the number of instances with instances.length and the number of feature dimensions with instances[i].length where i is an integer array index. How to test We will test your program on multiple test cases, where the datasets are generated by word2vec, a tool that is able to compute the feature vector of each word in the training corpus.1 Our datasets are first trained with a piece of corpus2 , then we further randomly sample a number of words from the whole 1https://code.google.com/p/word2vec/ 2http://mattmahoney.net/dc/text8.zip CS 540-1 Fall 2014 vocabulary and use their 200-dimensional feature vectors as our testing datasets. Your code should be able to handle test set of size 10,000, although we may also test it with smaller test sets. We will also vary centroid initialization, number of clusters, as well as threshold value, to test your program’s correctness comprehensively. The format of testing command will like this: java HW1 data file init centroid file threshold output flag where data file contains the features of all instances, init centroid file gives the initial positions of certains of number of centroids, and threhold has same meaning as mentioned. Besides, the outputFlag is an integer argument, controlling what the program outputs: • 1 – The array of centroids • 2 – The array of assignments • 3 – The array of distortions In order to facilitate your debugging, we will provide you a sample input and output files. They are words0.txt and initCentroid0.txt for the input, and output0 1.txt, output0 2.txt, output0 3.txt for the output, as seen in the zip file. So here is an example command: java HW1 words0.txt initCentroid0.txt 1e-4 3 You are NOT responsible for any file input or console output. We have written the class HW1 for you, which will load the data and pass it to the method you are implementing. As part of our testing process, we will unzip the file you return to us, remove any class files, call javac *.java to compile your code, and call the main method of HW1 with parameters of our choosing. Deliverables 1. Handin (use the handin program, see course webpage for details) your modified version of the code skeleton we provide you with your implementation of the k-means clustering algorithm. 2. Optionally, in the written part of your homework, add any comments about the program that you would like us to know.
