Presentation is loading. Please wait.

Presentation is loading. Please wait.

Data mining and statistical learning - lecture 12 Neural networks (NN) and Multivariate Adaptive Regression Splines (MARS)  Different types of neural.

Similar presentations


Presentation on theme: "Data mining and statistical learning - lecture 12 Neural networks (NN) and Multivariate Adaptive Regression Splines (MARS)  Different types of neural."— Presentation transcript:

1 Data mining and statistical learning - lecture 12 Neural networks (NN) and Multivariate Adaptive Regression Splines (MARS)  Different types of neural networks  Considerations in neural network modelling  Multivariate Adaptive Regression Splines

2 Data mining and statistical learning - lecture 12 Feed forward neural network Feed-forward neural network Input layer Hidden layer(s) Output layer x1x1 x2x2 xpxp z1z1 z2z2 zMzM … … f1f1 fKfK …

3 Data mining and statistical learning - lecture 12 Terminology Feed-forward network –Nodes in one layer are connected to the nodes in next layer Recurrent network –Nodes in one layer may be connected to the ones in previous layer or within the same layer

4 Data mining and statistical learning - lecture 12 Multilayer perceptrons Any number of inputs Any number of outputs One or more hidden layers with any number of units. Linear combinations of the outputs from one layer form inputs to the following layers Sigmoid activation functions in the hidden layers. x1x1 x2x2 xpxp z1z1 z2z2 zMzM … … f1f1 fKfK …

5 Data mining and statistical learning - lecture 12 Parameters in a multilayer perceptron C 1, C 2 : combination function g,  : activation function  0m  0k : bias of hidden unit  im  jk : weight of connection

6 Data mining and statistical learning - lecture 12 Least squares fitting of neural networks Consider a simple perceptron (no hidden layer) Find weights and bias minimizing the error function x1x1 x2x2 xpxp f1f1 f2f2 fKfK …

7 Data mining and statistical learning - lecture 12 Alternative measures of fit For regression we normally use the sum-of-squared errors as measure of fit For classification we use either squared errors or cross-entropy (deviance) and the corresponding classifier is argmax k f k (x) The measure of fit can also be adapted to specific distributions, such as Poisson distributions

8 Data mining and statistical learning - lecture 12 Combination and activation functions Combination function –Linear combination: –Radial combination: Activation function in the hidden layer –Identity –Sigmoid Activation function in the output layer –Softmax –Identity

9 Data mining and statistical learning - lecture 12 Ordinary radial basis function networks (ORBF) Input and output layers and one hidden layer Hidden layer: Combination function=radial Activation function=exponential, softmax Output layer: Combination function=linear Activation function =any, normally identity x1x1 x2x2 xpxp z1z1 z2z2 zMzM … … f1f1 fKfK …

10 Data mining and statistical learning - lecture 12 Issues in neural network modelling Preliminary training – learning with different initial weights (since multiple local minima are possible) Scaling of the inputs is important (standardization) The number of nodes in the hidden layer(s) The choice of activation function in the output layer –Interval – identity –Nominal – softmax

11 Data mining and statistical learning - lecture 12 Overcoming over-fitting 1.Early stopping 2.Adding a penalty function Objective function=Error function+Penalty term

12 Data mining and statistical learning - lecture 12 MARS: Multivariate Adaptive Regression Splines An adaptive procedure for regression that can be regarded as a generalization of stepwise linear regression

13 Data mining and statistical learning - lecture 12 Reflected pair of functions with a knot at the value x 1

14 Data mining and statistical learning - lecture 12 Reflected pairs of functions with knots at the values x 1 and x 2 x1x1 x2x2

15 Data mining and statistical learning - lecture 12 MARS with a single input X taking the values x 1, …, x N Form the collection of base functions Construct models of the form where each h m (X) is a function in C or a product of two or more such functions

16 Data mining and statistical learning - lecture 12 MARS model with a single input X taking the values x 1, x 2 x1x1 x2x2

17 Data mining and statistical learning - lecture 12 MARS model with a single input X taking the values x 1, x 2 x1x1 x2x2

18 Data mining and statistical learning - lecture 12 MARS: Multivariate Adaptive Regression Splines At each stage we consider as a new basis function pair all products of functions already in the model with one of the reflected pairs in the set C Although each basis function depends only on a single X j it is considered as a function over the entire input space

19 Data mining and statistical learning - lecture 12 MARS: Multivariate Adaptive Regression Splines - model selection MARS functions typically overfit the data and so a backward deletion procedure is applied The size of the model is determined by Generalized Cross Validation An upper limit can be set on the order of interaction

20 Data mining and statistical learning - lecture 12 The MARS model can be viewed as a generalization of the classification and regression tree (CART)

21 Data mining and statistical learning - lecture 12 Some characteristics of different learning methods CharacteristicNeural networks TreesMARS Natural handling of data of “mixed” typePoorGood Handling of missing valuesPoorGood Robustness to outliers in input spacePoorGoodPoor Insensitive to monotone transformations of inputsPoorGoodPoor Computational scalability (large N)PoorGood Ability to deal with irrelevant inputsPoorGood Ability to extract linear combinations of featuresGoodPoor InterpretabilityPoorFairGood Predictive powerGoodPoorFair

22 Data mining and statistical learning - lecture 12 Separating hyperplane

23 Data mining and statistical learning - lecture 12 Optimal separating hyperplane - support vector classifier margin Find the hyperplane that creates the biggest margin between the training points for class 1 and -1

24 Data mining and statistical learning - lecture 12 Formulation of the optimization problem Signed distance to decision border y=1 for one of the groups and y=-1 for the other one

25 Data mining and statistical learning - lecture 12 Two equivalent formulations of the optimization problem

26 Data mining and statistical learning - lecture 12 Characteristics of the support vector classifier Points well inside their class boundary do not play a big role in the shaping of the decision border Cf. linear discriminant analysis (LDA) for which the decision boundary is determined by the covariance matrix of the class distributions and their centroids

27 Data mining and statistical learning - lecture 12 Support vector machines using basis expansions (polynomials, splines)

28 Data mining and statistical learning - lecture 12 Characteristics of support vector machines The dimension of the enlarged feature space can be very large Overfitting is prevented by a built-in shrinkage of beta coefficients Irrelevant inputs can create serious problems

29 Data mining and statistical learning - lecture 12 The SVM as a penalization method Misclassification: f(x) 0 when y=-1 Loss function: Loss function + penalty:

30 Data mining and statistical learning - lecture 12 The SVM as a penalization method Minimizing the loss function + penalty is equivalent to fitting a support vector machine to data The penalty factor is a function of the constant providing an upper bound of

31 Data mining and statistical learning - lecture 12 Some characteristics of different learning methods CharacteristicNeural networks Support vector machines TreesMARS Natural handling of data of “mixed” typePoor Good Handling of missing valuesPoor Good Robustness to outliers in input spacePoor GoodPoor Insensitive to monotone transformations of inputs Poor GoodPoor Computational scalability (large N)Poor Good Ability to deal with irrelevant inputsPoor Good Ability to extract linear combinations of featuresGood Poor InterpretabilityPoor FairGood Predictive powerGood PoorFair


Download ppt "Data mining and statistical learning - lecture 12 Neural networks (NN) and Multivariate Adaptive Regression Splines (MARS)  Different types of neural."

Similar presentations


Ads by Google