CSE474-574-Assignment 4 Solved

Part 3.1 - Sparsity 
In this section we will interpret the coefficients from the final model you trained on all of the training data.

•   3.1.1 Count the number of non-zeros in each row of the train_vec matrix.

•   3.1.2 What is the average number non zero elements in each row?

•   3.1.3 On average what percentage of elements in each row have non-zero elements?

Part 3.2 - SVD 
Write out the result to a file called part_1.4_results.csv and submit this along with your assignment. (10 points) (You do not need to submit anything for your report for this part.)

• 3.2.1. What portion of the variance in your dataset is explained by each of the SVD dimensions?

Part 3.4 - Visualization 
•   3.4.1. Based on your observation, what is the difference between SVD and UMAP embeddings? 1-2 sentences should suffice.

•   3.4.2. Which one do you prefer to use for a classification task? why? 1-2 sentences should suffice

Part 4.1 - Clustering and evaluation
•   4.1.1 What is the range of possible values of silhouette coefficients?

•   4.1.2 Describe what a silhouette score of -1 and 1 mean?

•   4.1.3. Use silhouette score and KMeans from sklearn library to find the optimum number of clusters in your train_umap. Don’t forget to use SEED as your kmeans random_seed. In order to do this try different values of cluster numbers from 5 to 20. Choose the one that results in the best score.

•   4.1.4. Plot silhouette score for different values of n_clusters (a plot with n_clusters on the x-axis and silhouette score on the y-axis). Don’t forget to put the plot in your report.

Part 4.2 - Making a Kmeans classifier 

•   4.2.1 show your mapping (resulted dictionary) inside your project report.

Part 4.3 - Analyzing clusters (12 points)
•   4.3.1. Are there any two clusters in your clustering output with the same original label (for example, are there two clusters which both have same training label)? Use your visualizations and describe why?

•   4.3.2. Write the function bellow that returns nearest samples to a cluster center. Use this function and explain why there are overlaps in your labels?

•   4.3.3. Can you infere the overlapping label(s) by checking out most central samples? check with original labels.

Part 4.4 - Evaluate your Kmeans model on test dataset (12 points)
•   4.4.1. Using the generated mapping, and your clustering model, predict the labels of test dataset (you can use the embeddings of test data that you generated by umap test_umap)

•   4.4.2. Calculate the accuracy of model

•   4.4.3. Calculate both micro and macro values of precision, recall and F1 score

574   ONLY Part 5.1 - KNN classification
•   5.1.1. Train two seperate KNN models on both SVD and UMAP embeddings. Use n_neighbors=100.

•   5.1.2. Evaluate your model on test datas (test_umap and test_svd). Which model performs better? Why?

•   5.1.3. Calculate macro and micro precision recall and fscore for test_umap. Which one of the two do you prefer for evaluating your model? why?

•   5.1.4. Shortly describe why the two sets of values (macro and micro) are so similar in this case.

Contribution Statement (Minus 15 points if you do not submit this)
Please describe what each group member contributed to this project. Note that the professor and TA reserve the right to challenge this statement, and falsification of effort will be considered a violation of Academic Integrity . That is, if we find reason to believe that a specific group member’s claim about the work they contributed is not valid, then we reserve the right to take steps to ensure that the group member did, in fact, contribute as stated. We also reserve the right to adjust grades based on extreme differences in effort put into the assignment.