Machine Learning ICS 273AWhat is Expected?SyllabusMachine Learning according toSome ExamplesWhy is this cool/important?Types of LearningIngredientsSupervised Learning I1 nearest neighbors (your first ML algorithm!)1NN Decision SurfaceDistance MetricRemarks on NN methodsNon-parametric MethodsLogistic Regression / PerceptronThe logit / sigmoidObjectiveAlgorithm in detailA Note on Stochastic GDParametric MethodsHypothesis SpaceInductive BiasGeneralizationSlide 24Slide 25Slide 26Slide 27Slide 28Cross-validationMachine LearningICS 273AInstructor: Max WellingWhat is Expected?•Class•Homework (20%)•A Project (30%)•Final (50%)(subject to change – depending on availability of a reader)Programming in MATLAB.Syllabus•introduction: overview, examples, goals, algorithm evaluation, statistics.•Classification I: decision trees, random forests, boosting, k-nearest neighbors.•Classification 2: neural networks: perceptron, logistic regression, multi-layer networks, back- propagation.•Clustering & dimensionality reduction: k-means, expectation-maximization, PCA.•classification 3: kernel methods & support vector machines.•week 9/10: project presentations.week 11: final exam.Machine Learningaccording to •The ability of a machine to improve its performance based on previous results.•The process by which computer systems can be directed to improve their performance over time. •Subspecialty of artificial intelligence concerned with developing methods for software to learn from experience or extract knowledge from examples in a database.•The ability of a program to learn from experience — that is, to modify its execution on the basis of newly acquired information. •Machine learning is an area of artificial intelligence concerned with the development of techniques which allow computers to "learn". More specifically, machine learning is a method for creating computer programs by the analysis of data sets. Machine learning overlaps heavily with statistics, since both fields study the analysis of data, but unlike statistics, machine learning is concerned with the algorithmic complexity of computational implementations. ...Some Examples• ZIP code recognition• Loan application classification • Signature recognition• Voice recognition over phone• Credit card fraud detection• Spam filter• Collaborative Filtering: suggesting other products at Amazone.com • Marketing• Stock market prediction• Expert level chess and checkers systems• biometric identification (fingerprints, DNA, iris scan, face)• machine translation• web-search• document & information retrieval• camera surveillance• robosoccer• and so on and so on...Why is this cool/important?• Modern technologies generate data at an unprecedented scale.• The amount of data doubles every year.“One petabyte is equivalent to the text in one billion books, yet many scientific instruments, including the Large Synoptic Survey Telescope, will soon be generating several petabytes annually”. (2020 Computing: Science in an exponential world: Nature Published online: 22 March 2006)• Computers dominate our daily lives• Science, industry, army, our social interactions etc.We can no longer “eyeball” the images captured by some satellitefor interesting events, or check every webpage for some topic.We need to trust computers to do the work for us.Types of Learning• Supervised Learning• Labels are provided, there is a strong learning signal.• e.g. classification, regression. • Semi-supervised Learning.• Only part of the data have labels. • e.g. a child growing up.• Reinforcement learning.• The learning signal is a (scalar) reward and may come with a delay.• e.g. trying to learn to play chess, a mouse in a maze.• Unsupervised learning• There is no direct learning signal. We are simply trying to find structure in data.• e.g. clustering, dimensionality reduction.We will be concerned with these topics in thi sclassIngredients• Data: • what kind of data do we have?• Prior assumptions:• what do we know a priori about the problem?• Representation:• How do we represent the data?• Model / Hypothesis space:• What hypotheses are we willing to entertain to explain the data?• Feedback / learning signal:• what kind of learning signal do we have (delayed, labels)?• Learning algorithm:• How do we update the model (or set of hypothesis) from feedback?• Evaluation:• How well did we do, should we change the model?Supervised Learning IExample: Imagine you want to classify versus Data: 100 monkey images and 200 human images with labels what is what.,,{ 0}, 1,...,100{ 1}, 1,...,200i ij jx y ix y j= == =rrwhere x represents the greyscale of the image pixels andy=0 means “monkey” while y=1 means “human”.Task: Here is a new image: monkey or human?1 nearest neighbors(your first ML algorithm!)Idea: 1. Find the picture in the database which is closest your query image.2. Check its label.3. Declare the class of your query image to be the same as that of the closest picture.queryclosest image1NN Decision Surfacedecision curveDistance Metric•How do we measure what it means to be “close”?•Depending on the problem we should choose an appropriate distance metric.Hamming distance:( , ) | | { discrete};Scaled EuclideanDistance:( , ) ( ) ( ) { .};n m n mTn m n m n mD x x x x xD x x x x A x x x cont= - == - - =r r r rr r r r r rRemarks on NN methods• We only need to construct a classifier that works locally for each query. Hence: We don’t need to construct a classifier everywhere in space.• Classifying is done at query time. This can be computationally taxing at a time where you might want to be fast. • Memory inefficient (you have to keep all data around).• Curse of dimensionality: imagine many features are irrelevant / noisy  distances are always large.• Very flexible, not many prior assumptions.• k-NN variants robust against “bad examples”.Non-parametric Methods• Non-parametric methods keep all the data cases/examples in memory. • A better name is: “instance-based” learning• As the data-set grows, the complexity of the decision surface grows.• Sometimes, non-parametric methods have some parameters to tune...• Very few assumptions (we let the data speak).Logistic Regression / Perceptron•Fits a soft decision boundary between the classes.1 dimension2 dimensions(your second ML