Machine Learning

100%

1-1 Grading Program

90%

2-1 Linear Regression via Normal Equations

2-2 learn-linreg-NormEq with Gaussian elimination and QR decomposition

2-0 Word Count Program

Question 2. Datasets 
(a) The 1985 Auto Imports Database
- import-85.names is the helper file which contains information regarding the dataset
- import-85.data is a CSV file with the actual data

90%

3-1 Linear Regression with Gradient Descent　

3-2 Step Length for Gradient Descent

Question 3. Datasets 
(a) Airfare and demand: 
# Note: target is price
(b) Wine Quality:
# Note: target is quality
(c) Parkisons Dataset:
# Note: target is total UPDRS

80%

4-1 Logistic Regression with stochastic gradient ascent algorithm. (Classification)

4-2 Newton Algorithm (Classification)

Question 4. Datasets 
Classification dataset:
(a) Tic Tac Toe

100%

5-1 Ridge Regression using mini-Batch Gradient Descent algorithm (SGD)　

5-2 Learning Rate: AdaGrad, Bold-Driver and fixed stepsize　

5-3 Grid search with k-fold cross-validation for model selection　

Question 5. Datasets 
Classification datasets:
(a) Bank Marketing / bank.csv(https://archive.ics.uci.edu/ml/datasets/Bank+Marketing)
Regression datasets:
(b) Wine Quality(http://archive.ics.uci.edu/ml/datasets/Wine+Quality)

90%

6-1 Ordinary Least Squares with Stochastic Gradient Descent (SGD)

6-2 Ridge Regression with Stochastic Gradient Descent (SGD)

6-3 LASSO with Stochastic Gradient Descent (SGD)

6-4 Hyperparameter with GridSearchCV

6-5 Evaluate each model using cross val score

6-6 Higher Order Polynomial Regression

6-7 Lasso Regression along with Coordinate Descent.

Question 6. Datasets 
(a) Generate a Sample dataset called D1 :
i. Initialize matrix using Uniform distribution with μ = 1 and σ = 0.05
ii. Generate target using y = 1.3x^2 + 4.8x + 8 + ψ, where ψ randomly initialized.
(b) Wine Quality called D2: 
Winequality-red.csv (http://archive.ics.uci.edu/ml/datasets/Wine+Quality)

100%

7-1 K-Nearest Neighbor

7-2 Determine Optimal Value of K in KNN algorithm

Question 7. Datasets 
Classification Datasets: 
(a) Iris dataset D1: (https://archive.ics.uci.edu/ml/datasets/Iris)
# Note: Target attribute classes are Iris Setosa, Iris Versicolour and Iris Virginica 
(b) Wine Quality called D2: winequality-red.csv (http://archive.ics.uci.edu/ml/datasets/Wine+Quality)

8 NA

80%

9-1 Spam filter using SVM

9-2 Compare SVM based spam filter with another model

Question 9. Datasets
(a) Sparse Dataset / w8a dataset D1: (https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html#w8a)
(b) UCI Dataset / SMS Spam D2: (https://archive.ics.uci.edu/ml/datasets/SMS+Spam+Collection)
(c) UCI Dataset / Spambase D3: (https://archive.ics.uci.edu/ml/datasets/Spambase)

98%

10-1 Implement basic matrix factorization (MF) technique for recommender systems

10-2 Recommender Systems using matrix factorization libmf / sckit-learn

Question 10. Datasets 
Recommender Datasets:
(a) movielens 100k dataset D1: Rating prediction dataset (http://grouplens.org/datasets/movielens/100k/)
(b) The RMSE score for rating prediction is available at Mymedialite website (http://www.mymedialite.net/examples/datasets.html)

80%

11-1 K Means clustering algorithm

Question 10. Datasets 
Sparse dataset :
(a) IRIS dataset D1: (https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multiclass/iris.scale)
(b) rcv1v2 (topics; subsets) D2: (https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/multilabel.html)
(c) 20Newsgroups dataset D3: (http://qwone.com/~jason/20Newsgroups/)