Machine Learning Assignments of the course COL774 taken by Parag Singla, at IIT Delhi.
Assignment portion was way too time consuming, but the weightage was around (33 marks), which is very unbalanced for a rational person. Final exam has more weightage than this.
More than 100 hrs of work vs 2 hrs of major exam. Fancy that.
3/10 would recommend
Obviously not: Copying assignments will not give you the opportunity to learn, and hence it would defeat the purpose of the course; after saying this, I remember the meaningful quote of Picasso, "Good artists borrow, great artists steal." It's upto you, whether you want to be good or great. The separation line is absurdly wide but a fool never removes his eyes from the sky.
- Linear Regression
- Locally Weighted Linear Regression
- Logistic Regression
- Gaussian Discriminant Analysis
Correctly done
- Text Classification Using Naive Bayes
- MNIST Digit Classification using SVM
Correctly done
- Decision Trees and Random Forests
- Neural Networks
Demo: When pruning, one should consider the best possible node to prune rather going in leaf to root fashion: I didn't implement it with purpose: It took too much time to train.
- Competitive Part:
- SVM and CNN on a given dataset
- Non-competitive Part:
- Do any model on given dataset
Least interesting assignment: Test data was wrong for like a week or something, deadlines were changed way too quickly, it took way too much time to train the model: Ill-managed.
SVM and CNN are correctly implemented.