Machine Learning

Transcript

1. Machine Learning Andrea Iacono https://github.com/andreaiacono/MachineLearning
2. Machine Learning: Intro [Wikipedia]: a branch of artificial intelligence that allows the construction and the study of systems that can learn from data What is Machine Learning?
3. - Regression analysis - Similarity and metric learning - Decision tree learning - Association rule learning - Artificial neural networks - Genetic programming - Support vector machines (classification and regression analysis) - Clustering - Bayesian networks Machine Learning: Intro Some approaches:
4. Machine Learning: Intro Supervised learning vs Unsupervised learning Machine learning vs Data mining
5. Machine Learning: Regression analysis Regression Analysis A statistical technique for estimating the relationships among a dependent variable and independent variables
6. Machine Learning: Regression analysis Prediction of house prices Size (x) Price (y) 0.80 70 0.90 83 1.00 74 1.10 93 1.40 89 1.40 58 1.50 85 1.60 114 1.80 95 2.00 100 2.40 138 2.50 111 2.70 124 3.20 172 3.50 172
7. Machine Learning: Regression analysis Prediction of house prices Hypothesis: h θ (x)=θ0 +θ1 x
8. Machine Learning: Regression analysis Prediction of house prices Hypothesis: h θ (x)=θ0 +θ1 x J (θ 0, θ 1 )= 1 2m ∑ i=1 m (h θ (x(i))−y(i))2 Cost function for linear regression:
9. Machine Learning: Regression analysis Prediction of house prices Hypothesis: h θ (x)=θ0 +θ1 x J (θ 0, θ 1 )= 1 2m ∑ i=1 m (h θ (x(i))−y(i))2 Gradient Descent repeat until convergence : θ0 =θ0 −α 1 m ∑ i=1 m (h θ ( x(i))− y(i)) θ 1 =θ 1 −α 1 m ∑ i=1 m [(h θ (x(i))− y(i)) x(i) ] Cost function for linear regression:
10. Machine Learning: Regression analysis Prediction of house prices Iterative minimization of cost function with gradient descent
11. Machine Learning: Regression analysis Hands on
12. Machine Learning: Regression analysis Regression analysis - one / multiple variables - linear / higher order curves - several optimization algorithms - linear regression - logistic regression - simulated annealing - ...
13. Machine Learning: Regression analysis Overfitting vs underfitting
14. Machine Learning: Similarity and metric learning Similarity and metric learning - concept of distance
15. Euclidean distance euclidean distance (p ,q )= √∑ i =1 n (p i −q i )2 Machine Learning: Similarity and metric learning
16. manhattan distance(p ,q )=∑ i=1 n ∣(p i −q i )∣ Machine Learning: Similarity and metric learning Manhattan distance
17. Pearson ' s correlation (p ,q )= ∑ i =1 n (p i q i )− ∑ i =1 n p i ∑ i =1 n q i n √(∑ i =1 n p i 2 − (∑ i =1 n p i ) 2 n )(∑ i =1 n q i 2 − (∑ i =1 n q i ) 2 n ) Machine Learning: Similarity and metric learning Pearson's correlation
18. Searches a large group of users for finding a small subset that have tastes like yours. Based on what this subset likes or dislikes the system can recommend you other items. Two main approaches: - User based filtering - Item based filtering Machine Learning: Similarity and metric learning Collaborative filtering
19. Machine Learning: Similarity and metric learning User based filtering - based on ratings given to the items, we can measure the distance among users - we can recommend to the user the items that have the highest ratings among the closest users
20. Hands on Machine Learning: Similarity and metric learning
21. Machine Learning: Similarity and metric learning Is user based filtering good for - scalability? - sparse data? - quickly changing data?
22. Machine Learning: Similarity and metric learning Is user based filtering good for - scalability? - sparse data? - quickly changing data? No, it's better to use item based filtering
23. Machine Learning: Similarity and metric learning Euclidean distance for item based filtering: nothing has changed! - based on ratings got from the users, we can measure the distance among items - we can recommend an item to a user, getting the items that are closer to the highest rated by the user
24. Hands on Machine Learning: Similarity and metric learning
25. P (A∣B )= P (B∣A)P (A) P (B ) Machine Learning: Bayes' classifier Example: given a company where 70% of developers use Java and 30% use C++, and knowing that half of the Java developers always use enhanced for loop, if you look at the snippet: which is the probability that the developer who wrote it uses Java? for (int j=0; j<100; j++) { t = tests[j]; } P (A∣B )= P (B∣A)P (A) P (B ) Bayes' theorem
26. P (A∣B )= P (B∣A)P (A) P (B ) Machine Learning: Bayes' classifier Example: given a company where 70% of developers use Java and 30% use C++, and knowing that half of the Java developers always use enhanced for loop, if you look at the snippet: which is the probability that the developer who wrote it uses Java? for (int j=0; j<100; j++) { t = tests[j]; } P (A∣B )= P (B∣A)P (A) P (B ) Bayes' theorem Hint: A = developer uses Java B = developer writes old for loops
27. P (A∣B )= P (B∣A)P (A) P (B ) Machine Learning: Bayes' classifier Example: given a company where 70% of developers use Java and 30% use C++, and knowing that half of the Java developers always use enhanced for loop, if you look at the snippet: which is the probability that the developer who wrote it uses Java? for (int j=0; j<100; j++) { t = tests[j]; } P (A∣B )= P (B∣A)P (A) P (B ) Bayes' theorem Solution: A = developer uses Java B = developer writes old for loops P (A∣B)= P (B∣A)P (A) P (B) = 0.5⋅0.7 0.65 =0.54 P(A) = prob. that a developer uses Java = 0.7 P(B) = prob. that any developer uses old for loop = 0.3 + 0.7*0.5 = 0.65 P(B|A) = prob. that a Java developer uses old for loop = 0.5
28. Machine Learning: Bayes' classifier Naive Bayes' classifier - supervised learning - trained on a set of known classes - computes probabilities of elements to be in a class - smoothing required P c (w 1 , .... ,w n )= ∏ i =1 n P (c∣w i ) ∏ i =1 n P (c∣w i )+∏ i=1 n (1−P (c∣w i ))
29. Machine Learning: Bayes' classifier Naive Bayes' classifier Example - we want a classifier for Twitter messages - define a set of classes: {art, tech, home, events,.. } - trains the classifier with a set of alreay classified tweets - when a new tweet arrives, the classifier will (hopefully) tell us which class it belongs to
30. Machine Learning: Bayes' classifier Hands on
31. Machine Learning: Bayes' classifier Sentiment analysis - define two classes: { +, - } - define a set of words: { like, enjoy, hate, bore, fun, …} - train a NBC with a set of known +/- comments - let NBC classify any new comment to know if +/- - performance is related to quality of training set
32. Machine Learning: Clustering - Unsupervised learning - Different algorithms: - Hierarchical clustering - K-Means clustering - ... Clustering Common use cases: - navigation habits - online commerce - social/political attitudes - ...
33. Machine Learning: Clustering K-Means aims at identifying cluster centroids, such that an item belonging to a cluster X, is closer to the centroid of cluster X than to the centroid of any other cluster. K-Means clustering
34. Machine Learning: Clustering The algorithm requires a number of clusters to start, in this case 3. The centroids are placed in the item space, typically in random locations. K-Means clustering
35. Machine Learning: Clustering The algorithm will then assign to each centroid all items that are closer to it than to any other centroid. K-Means clustering
36. Machine Learning: Clustering The centroids are then moved to the center of mass of the items in the clusters. K-Means clustering
37. Machine Learning: Clustering A new iteration occurs, taking into account the new centroid positions. K-Means clustering
38. Machine Learning: Clustering The centroids are again moved to the center of mass of the items in the clusters. K-Means clustering
39. Machine Learning: Clustering Another iteration occurs, taking into account the new centroid positions. K-Means clustering
40. Machine Learning: Clustering The centroids are again moved to the center of mass of the items in the clusters. K-Means clustering
41. Machine Learning: Clustering Another iteration occurs, taking into account the new centroid positions. Note that this time the cluster membership did not change. The cluster centers will not move anymore. K-Means clustering
42. Machine Learning: Clustering The solution is found. K-Means clustering
43. Machine Learning: Clustering Hands on
44. Machine Learning: Neural networks A logical calculus of the ideas immanent in nervous activity by McCulloch and Pitts in 1943 Neural networks
45. Machine Learning: Neural networks Neural networks Feedforward Perceptron
46. Machine Learning: Neural networks Neural networks Logic operators with neural networks: Threshold = 0 X0 X1 X2 Σ Result -10 0 0 -10 0 -10 0 20 10 1 -10 20 0 10 1 -10 20 20 30 1 OR operator
47. Machine Learning: Neural networks Neural networks Threshold = 0 X0 X1 X2 Σ Result -30 0 0 -30 0 20 -30 20 0 -30 20 20 which operator? Logic operators with neural networks:
48. Machine Learning: Neural networks Neural networks Threshold = 0 X0 X1 X2 Σ Result -30 0 0 -30 0 -30 0 20 -10 0 -30 20 0 -10 0 -30 20 20 10 1 AND operator Logic operators with neural networks:
49. Machine Learning: Neural networks Hands on
50. Machine Learning: Neural networks Neural networks Backpropagation Phase 1: Propagation - Forward propagation of a training pattern's input through the neural network in order to generate the propagation's output activations - Backward propagation of the propagation's output activations through the neural network using the training pattern target in order to generate the deltas of all output and hidden neurons Phase 2: Weight update - Multiply its output delta and input activation to get the gradient of the weight - Bring the weight in the opposite direction of the gradient by subtracting a ratio of it from the weight
51. Machine Learning: Neural networks Neural networks Multilayer perceptrons
52. Machine Learning: Neural networks Hands on
53. Machine Learning: Genetic algorithms Genetic algorithms GA is a programming technique that mimics biological evolution as a problem-solving strategy Steps - maps the variables of the problem into a sequence of bits, a chromosome Chromosome - creates a random population of chromosomes - let evolve the population using evolution laws: - the higher the fitness, the higher the chance of breeding - crossover of chromosomes - mutation in chromosomes - if otpimal solution is found or after n steps the process is stopped
54. Machine Learning: Genetic algorithms Genetic algorithms Mutation Crossover
55. Hands on Machine Learning: Genetic algorithms
56. Thanks! Machine Learning The code is available on: https://github.com/andreaiacono/MachineLearning