TODO

Put readings on todo list can be found on class webpage Put homeworks on todo list Midterm 1 Monday march 16 in class Discussion sections Monday-Wednesday

Core material

• Find patterns in data and use them to make predictions
• Models + stats help us understand patterns
• Use optimization algorithms to learn the patterns

Classification

• Simplest Case: you have two choices, given data, make a prediction
• Knn best when few outliers

Classifying Numbers

1. Turn points into grid of 0s and 1s based on the color of the grid
2. Turn grid into a vector by flattening the vector
3. Create hyperplane in the n-dimensional space to group things

Testing and Validation

• Train a classifier - it learns to distinguish 7 from not 7
• Test the classifier on NEW images

• There are two types of error

  • Training set error: Fraction of training images not classified correctly
  • Test set error: Fraction of misclassified NEW images, not seen during training
• Outliers: Points whose labels are atypical (e.g solvent borrowers who defaulted anyway)
• Overfitting: When the test error deteriorates because the classifier becomes too sensitive to outliers
• Hyperparameters: Most ML algorithms have a few hyperparameters that control over/underfitting. eg k in k-nearest neighbors

Select classifiers by validation

• Validation Set: Hold back a subset of the labeled data
• Train the classifier multiple times with different hyperparameter settings
• Choose setting(hyperparameter + learning algorithm) that works best on validation set

Now, we have 3 sets:

• Training set: Used to learn model weights
• Validation set: Used to tune hyperparameters, choose among different models
• Test set: Used as FINAL evaluation. Test set kept in vault, ran once, at the very end

Kaggle.com

• Runs ML competitions, including our HWs

• We use 2 data sets:

  • public set labels available during the competition
  • private set labels known only to Kaggles

Techniques of Machine Learning Taught in Class

Supervised learning

• Classification: Is this email spam?
• Regression: How likely does this patient have cancer?

Unsupervised learning

• Clustering: which DNA sequences are similar to each other?
• Dimensionality Reduction: What are common features of faces?

Questions