Homework answers / question archive / Please help me out on this MATLAB Lab Assignment titled "Distance based clustering and classification" using concepts of Linear Algebra

Please help me out on this MATLAB Lab Assignment titled "Distance based clustering and classification" using concepts of Linear Algebra

Computer Science

Share With

---

Please help me out on this MATLAB Lab Assignment titled "Distance based clustering and classification" using concepts of Linear Algebra.
 

7.2 CLUSTERING AND CLASSIFICATION

 

7.2.1 Clustering

 

In machine learning applications, the goal is to use data to learn how to automate certain

tasks, e.g., image classification, identity verification, disease risk estimation, stock market pre-

diction, etc. A key tool for working with data is Clustering also known as unsupervised learning

or community detection. A cluster is defined as a group of similar objects, using a similarity

metric appropriate for the application at hand. You can think of similarity as moving in

the opposite direction of distance: if distance increases, similarity decreases. Clustering then

refers to algorithms and techniques to group (or cluster) objects according to their similarity.

Examples include

 

1. Community detection in social networks. Users form clusters based on political view, socio-

economic background, location, age, etc.

 

2. Computer vision. Images can be grouped according to similarity, e.g., to search images representing specific objects on the Internet. Video analysis can be used for pedestrian/-

car detection in self driving cars.

 

3. Genetics’: grouping genes according to common characteristics.

 

4. Marketing: group consumers according to behavior and spending habits. Identify new

trends by detecting emerging new communities.

 

Clustering presents several major challenges

 

¢ We may not know a priori how many clusters there will be.

 

e Similarity may be subjective or it may be difficult to find a reliable metric to quantify similarity.

 

e Clusters may evolve over time.

 

7.2.2 Classification

Clustering can also be interpreted as a method to understand data. Once the clusters are known, they can be used to process new data that comes in. For example:

11/14/22, 8:04 PM Submit Lab 7 | Gradescope https://www.gradescope.com/courses/417083/assignments/2308594/submissions/new 1/6 0/11 Questions Answered Lab 7 Q1 10 Points Submit one image for each image type (label). For example you can display the image at index 1 using: img = reshape(data(1,:),28,28); imshow(img); Image Upload - JPG Only: Please select file(s) Select file(s) Save Answer Q2 30 Points Q2.1 10 Points Report the cluster center as an image for . You can do this by displaying each row of as an image, then saving that image using . Image Upload - JPG Only: Please select file(s) Select file(s) Code SnippetN: Choose Files o file chosen ö ci i = 1, 2, 3 C saveas(gcf, "cluster_1.jpg") Choose Files No file chosen ö Choose Files No fififile chosen 11/14/22, 8:04 PM Submit Lab 7 | Gradescope https://www.gradescope.com/courses/417083/assignments/2308594/submissions/new 2/6 Enter your answer here Save Answer Q2.2 10 Points Compute the error rate using the equation: Explanation: Provide the error rate here Enter your answer here Code Snippet: Enter your answer here Save Answer Q2.3 10 Points Give 3 examples, one for each class, where the assigned cluster is wrong. Submit the image, image index (which row of data this is), and the assigned cluster. Try to explain the possible reason(s) why they are wrongly clustered. Choose Files No fifile chosen error rate = ∗ #number of data points #data points in incorrect cluster 100. Choose Files No fififile chosen 11/14/22, 8:04 PM Submit Lab 7 | Gradescope https://www.gradescope.com/courses/417083/assignments/2308594/submissions/new 3/6 Explanation: Indices and which cluster they were mistakenly assigned to: Enter your answer here Image Upload - JPG Only: Please select file(s) Select file(s) Code Snippet: Enter your answer here Save Answer Q3 30 Points Q3.1 10 Points Implement Algorithm 2 for all points in the testing data. Report the classification results for the first 5 samples (i.e which label it is assigned). Explanation: Indices and which cluster they were assigned to: Enter your answer here Code Snippet: Choose Files No fififile chosen ö Choose Files No fifile chosen 11/14/22, 8:04 PM Submit Lab 7 | Gradescope https://www.gradescope.com/courses/417083/assignments/2308594/submissions/new 4/6 Enter your answer here Save Answer Q3.2 10 Points Report the classification error for all testing samples with the true test labels in . Explanation Error Rate for Test Data: Enter your answer here Code Snippet: Enter your answer here Save Answer Q3.3 10 Points Give 3 examples of wrongly classified images the test data, one for each label. Explanation: Indices and which cluster they were assigned to: Choose Files No fififile chosen labels_test Choose Files No fifile chosen 11/14/22, 8:04 PM Submit Lab 7 | Gradescope https://www.gradescope.com/courses/417083/assignments/2308594/submissions/new 5/6 Enter your answer here Image Upload - JPG Only: Please select file(s) Select file(s) Save Answer Q4 24 Points Q4.1 8 Points What is the minimum number of clusters one should choose when performing clustering methods like k-means on a new dataset? Enter your answer here Save Answer Q4.2 8 Points How would you use distance to determine how similar digits are to each other? Does smaller distance means similar digits? Enter your answer here Save Answer ö 11/14/22, 8:04 PM Submit Lab 7 | Gradescope https://www.gradescope.com/courses/417083/assignments/2308594/submissions/new 6/6 Q4.3 8 Points How would you use the result to determine which of the digits have more variation in terms of how they are written? Enter your answer here Save Answer Q5 Code Submission 6 Points Submit your run_kmeans.m file here: Please select file(s) Select file(s) Save Answer Save All Answers Submit & View Submission T ö