1. Machine Learning Week 1, Lecture 2

2. Recap Supervised Learning Data Set Learning Algorithm Hypothesis h h(x) ≈ f(x) Unknown Target f Hypothesis Set 5 0 4 1 9 2 1 3 1 4 Hyperplane Halfspace >0 Halfspace < 0 w ClassificationRegression

3. np-hard in general Assume Data Is Linear Separable!!! Perception find separating hyperplane Convex

4. Today Convex Optimization – Convex sets – Convex functions Logistic Regression – Maximum Likelihood – Gradient Descent Maximum likelihood and Linear Regression

5. Convex Optimization Optimization problem, in general very hard (if possible at all)!!! For convex optimization problems theoretical (polynomial time) and practical solutions exist (most of the time) Example:

6. Convex Sets Convex Set Non-convex Set The “line” from x to y must also be in the set

7. Convex Sets Union of convex setsmay not be convex Intersection of convex sets

8. Convex Functions x,f(x) y,f(y) f is concave if –f is convex Concave?, Convex? Both

9. Differentiable Convex Functions x,f(x) y,f(y) f(x)+f’(x)(y-x) Example

10. Twice Differentiable Convex Functions f is convex if the Hessian is positive semi-definite for all x. Real symmetric matrix A is positive semidefinite if for all nonzero x 1D:

11. Simple 2D Example

12. More Examples Quadratic Functions: Convex if A is positive semidefinite Affine Functions:

13. Convexity of Linear Regression Quadratic Functions: Convex if A is positive semidefinite Real and Symmetric:Clearly

14. Epigraph Connection between convex sets and convex functions f is convex if epi(f) is a convex set

15. Sublevel sets Convex function Define α-Sublevel set: Is Convex

16. Convex Optimization f and g are convex, h is affine Local Minima are Global Minima

17. Examples of Convex Optimization Linear Programming Quadratic Programming (P is positive semidefinite)

18. Summary Rockafellar stated, in his 1993 SIAM Review survey paper “In fact the great watershed in optimization isn’t between linearity and nonlinearity, but convexity and nonconvexity” Convex GOOD!!!!

19. Estimating Probabilities Probability of getting cancer given your situation. Probability that AGF wins against Viborg given the last 5 results. Probability that the loan is not payed back as a function of credit worthiness Probability of a student getting an A in Machine Learning given his grades. Data is actual events not probabilities, e.g. some students that failed and some who did not…

20. Breast Cancer http://archive.ics.uci.edu/ml/datasets/Breast+Cancer+Wisconsin+%28Original%29 1. Sample code number: id number 2. Clump Thickness: 1 - 10 3. Uniformity of Cell Size: 1 - 10 4. Uniformity of Cell Shape: 1 - 10 5. Marginal Adhesion: 1 - 10 6. Single Epithelial Cell Size: 1 - 10 7. Bare Nuclei: 1 - 10 8. Bland Chromatin: 1 - 10 9. Normal Nucleoli: 1 - 10 10. Mitoses: 1 - 10 Input Features benign malignant Target Function PREDICT PROBABILITY OF BENIGN AND MALIGNANT ON FUTURE PATIENTS

21. Maximum Likelihood Biased Coin, (bias θ probability of heads) Flip it n times independently (Bernoulli trials), Count the number of heads k Fix θ, What is the probability of seeing D Take Logs After seeing the data what can we infer Likelihood of the data

22. Maximum Likelihood solve for 0 Compute Gradient Negative Log Likelihood of the data (log is monotone) Maximize Minimize

23. Bayesian Perspective Bayes Rule: Want: Need: A Prior Likelihood x Prior Normalizing factor Posterior

24. Bayesian Perspective Compute the probability of each hypotheses Pick the most likely and use for predictions (map = maximum a posteriori) Compute Expected Values (Weighed average over all hypotheses)

25. Logistic Regression Assume Independent Data Points, Apply Maximum Likelihood (there is a Bayesian version to) Hard Threshold Hard and Soft Threshold Can and is used for classification. Predict most likely y

26. Maximum Likelihood Logistic Regression Neg. Log likelihood is convex Cannot solve for zero analytically

27. Descent Methods Iteratively move toward a better solution Numerically we are doing small pertubation or mutations if you want in each variable e.g. O(dim * time(eval cost function)) time Show contour plot where f is twice continuously differentiable Pick start point x Repeat Until Stopping Criterion Satisfied Compute Descent Direction v Line Search: Compute Step Size t Update: x = x + t v Gradient Descent

28. Line (Ray) Search Pick start point x Repeat Until Stopping Criterion Satisfied Compute Descent Direction v Line Search: Compute Step Size t Update: x = x + t v Solve analytically (if possible) Backwards Search start high and decrease until improving distance found [SL 9.2] Fix to a small constant Use size of the gradient scaled with small constant. Start with constant, let it decrease slowly or when to high

29. Stopping Criteria Gradient becomes very small Max number of iterations used

30. Gradient Descent for Linear Reg.

31. GD For Linear Regression Matlab style function theta= GD(X,y,theta) LR = 0.1 for i=1:50 cost = (1/length(y))* sum((X*theta-y).^2) grad = (1/length(y))*2.*X'*(X*theta-y) theta = theta – LR * grad end Note we do not scale gradient to unit vector

32. Learning Rate

33. Gradient Descent Jump Around Use Exact Line Search Starting From (10,1)

34. Gradient Descent Running Time Number of iterations x Cost per iteration. Cost Per Iteration is usually not a problem. Number of iterations depends choice of line search and stopping Criterion clearly. – Very Problem and Data Specific – Need a lot of math to give bounds. – We will not cover it in this course.

35. Gradient Descent For Logistic Regression Handin 1! A long with multiclass extension

36. Stochastic Gradient Descent Pick at random and use Use K points chosen at random Mini-Batch Gradient Descent

37. Linear Classification with K classes Use Logistic regression All Vs one. – Train K classifiers one for each class – Input X is the same. Y is 1 for all elements from that class and 0 otherwise (All vs. One) – Prediction, compute the probability for all K classifiers output class with highest probability. Use Softmax Regression – Extension of logistic function to K classes in some sense – Covered in Handin 1.

38. Maximum Likelihood and Linear Regression (Time to spare slide) Assume: Independently

39. Todays Summary Convex Optimizations – Many Definitions – Local Optimal is Global Optimal – Usually theoretical and practically feasible Maximum likelihood – Use as a proxy for – Assume Independent Data Gradient Descent – Minimize function – Iteratively finding better solution by local steps based on gradient – First order method (Uses gradient) – Other methods exist, e.g. Second order methods (use hessian)