1©2005-2007 Carlos Guestrin1Neural NetworksMachine Learning – 10701/15781Carlos GuestrinCarnegie Mellon UniversityFebruary 14th, 2007©2005-2007 Carlos Guestrin2Logistic regression P(Y|X) represented by: Learning rule – MLE:2©2005-2007 Carlos Guestrin3Sigmoid-6 -4 -2 0 2 4 600.10.20.30.40.50.60.70.80.91-6 -4 -2 0 2 4 600.10.20.30.40.50.60.70.80.91-6 -4 -2 0 2 4 600.10.20.30.40.50.60.70.80.91w0=0, w1=1w0=2, w1=1 w0=0, w1=0.5©2005-2007 Carlos Guestrin4Perceptron as a graph-6 -4 -2 0 2 4 600.10.20.30.40.50.60.70.80.913©2005-2007 Carlos Guestrin5Linear perceptronclassification region-6 -4 -2 0 2 4 600.10.20.30.40.50.60.70.80.91©2005-2007 Carlos Guestrin6Optimizing the perceptron Trained to minimize sum-squared error4©2005-2007 Carlos Guestrin7Derivative of sigmoid©2005-2007 Carlos Guestrin8The perceptron learning rule Compare to MLE:5©2005-2007 Carlos Guestrin9Percepton, linear classification,Boolean functions Can learn x1 Ç x2 Can learn x1 Æ x2 Can learn any conjunction or disjunction©2005-2007 Carlos Guestrin10Percepton, linear classification,Boolean functions Can learn majority Can perceptrons do everything?6©2005-2007 Carlos Guestrin11Going beyond linear classification Solving the XOR problem©2005-2007 Carlos Guestrin12Hidden layer Perceptron: 1-hidden layer:7©2005-2007 Carlos Guestrin13Example data for NN with hidden layer©2005-2007 Carlos Guestrin14Learned weights for hidden layer8©2005-2007 Carlos Guestrin15NN for images©2005-2007 Carlos Guestrin16Weights in NN for images9©2005-2007 Carlos Guestrin17Forward propagation for 1-hiddenlayer - Prediction 1-hidden layer:©2005-2007 Carlos Guestrin18Gradient descent for 1-hidden layer –Back-propagation: ComputingDropped w0 to make derivation simpler10©2005-2007 Carlos Guestrin19Gradient descent for 1-hidden layer –Back-propagation: ComputingDropped w0 to make derivation simpler©2005-2007 Carlos Guestrin20Multilayer neural networks11©2005-2007 Carlos Guestrin21Forward propagation – prediction Recursive algorithm Start from input layer Output of node Vk with parents U1,U2,…:©2005-2007 Carlos Guestrin22Back-propagation – learning Just gradient descent!!! Recursive algorithm for computing gradient For each example Perform forward propagation Start from output layer Compute gradient of node Vk with parents U1,U2,… Update weight wik12©2005-2007 Carlos Guestrin23Many possible response functions Sigmoid Linear Exponential Gaussian …©2005-2007 Carlos Guestrin24Convergence of backprop Perceptron leads to convex optimization Gradient descent reaches global minima Multilayer neural nets not convex Gradient descent gets stuck in local minima Hard to set learning rate Selecting number of hidden units and layers = fuzzy process NNs falling in disfavor in last few years We’ll see later in semester, kernel trick is a good alternative Nonetheless, neural nets are one of the most used MLapproaches13©2005-2007 Carlos Guestrin25Training set error Neural nets representcomplex functions Output becomes more complexwith gradient steps©2005-2007 Carlos Guestrin26Overfitting Output fits training data “too well” Poor test set accuracy Overfitting the training data Related to bias-variance tradeoff One of central problems of ML Avoiding overfitting? More training data Regularization Early stopping14©2005-2007 Carlos Guestrin27What you need to know aboutneural networks Perceptron: Representation Perceptron learning rule Derivation Multilayer neural nets Representation Derivation of backprop Learning rule Overfitting Definition Training set versus test set Learning curve©2005-2007 Carlos Guestrin28Instance-basedLearningMachine Learning – 10701/15781Carlos GuestrinCarnegie Mellon UniversityFebruary 14th, 200715©2005-2007 Carlos Guestrin29Why not just use Linear Regression?©2005-2007 Carlos Guestrin30Using data to predict new data16©2005-2007 Carlos Guestrin31Nearest neighbor©2005-2007 Carlos Guestrin32Univariate 1-Nearest NeighborGiven datapoints (x1,y1) (x2,y2)..(xN,yN),where we assume yi=f(xi) for someunknown function f.Given query point xq, your job is to predictNearest Neighbor:1. Find the closest xi in our set of datapoints( )qxfy !ˆ( )qiixxnni !=argmin( )nniyy =ˆ2. PredictHere’s adataset withone input, oneoutput and fourdatapoints.xyHere, this isthe closestdatapointHere, this isthe closestdatapointHere, this isthe closestdatapointHere, this isthe closestdatapoint17©2005-2007 Carlos Guestrin331-Nearest Neighbor is an example of…. Instance-based learningFour things make a memory based learner: A distance metric How many nearby neighbors to look at? A weighting function (optional) How to fit with the local points?x1 y1x2 y2x3 y3..xn ynA function approximatorthat has been aroundsince about 1910.To make a prediction,search database forsimilar datapoints, and fitwith the local points.©2005-2007 Carlos Guestrin341-Nearest NeighborFour things make a memory based learner:1. A distance metricEuclidian (and many more)2. How many nearby neighbors to look at?One3. A weighting function (optional)Unused4. How to fit with the local points?Just predict the same output as the nearest neighbor.18©2005-2007 Carlos Guestrin35Multivariate 1-NN examplesRegression Classification©2005-2007 Carlos Guestrin36Multivariate distance metricsSuppose the input vectors x1, x2, …xn are two dimensional:x1 = ( x11 , x12 ) , x2 = ( x21 , x22 ) , …xN = ( xN1 , xN2 ).One can draw the nearest-neighbor regions in input space.Dist(xi,xj) =(xi1 – xj1)2+(3xi2 – 3xj2)2The relative scalings in the distance metric affect region shapes.Dist(xi,xj) = (xi1 – xj1)2 + (xi2 –
View Full Document