PDF4PRO ⚡AMP

Modern search engine that looking for books and documents around the web

Example: dental hygienist

Machine Learning: Generative and Discriminative Models

Machine Learning: Generative and Discriminative ModelsSargur N. Learning Course: ~srihari/CSE574 LearningSrihari2 Outline of Presentation1. What is Machine Learning?ML applications, ML as Search2. Generative and Discriminative Taxonomy3. Generative - Discriminative PairsClassifiers: Na ve Bayes and Logistic RegressionSequential Data: HMMs and CRFs4. Performance Comparison in Sequential ApplicationsNLP: Table extraction, POS tagging, Shallow parsing, Handwritten word recognition, Document analysis5. Advantages, disadvantages6. Summary7. ReferencesMachine LearningSrihari31. Machine Learning Programming computers to use example data or past experience Well-Posed Learning Problems A computer program is said to learn from experience E with respect to class of tasks T and performance measure P, if its performance at tasks T, as measured by P, improves with experience LearningSrihari4 Problems Too Difficult To Program by Hand Learning to drive an autonomous vehicle Train computer-controlled vehicles to steer correctly Drive at 70 mph for 90 miles on public highways Associate steering commands with image sequencesTask T: driving on pu

• Gaussians, Naïve Bayes, Mixtures of multinomials • Mixtures of Gaussians, Mixtures of experts, Hidden Markov Models (HMM) ... – by fitting Gaussian class-conditional densities will result in . 2M . parameters for means, M(M+1)/2 ... Markov Random Field (MRF)

Fullscreen Download
Loading..

Tags:

  Field, Mixtures, Random, Gaussian, Markov, Markov random field

Information

Domain:

Source:

Link to this page:

Please notify us if you found a problem with this document:

Spam in document Broken preview Other abuse

Transcription of Machine Learning: Generative and Discriminative Models

Documents from same domain

Gaussian Distribution - Welcome to CEDAR

cedar.buffalo.edu

• For a multivariate Gaussian distribution N(x| µ,Λ-1) for a D-dimensional variable x – Conjugate prior for mean µ assuming known precision is Gaussian – For known mean and unknown precision matrix Λ, conjugate prior is Wishart distribution – If both mean and precision are unknown conjugate prior is Gaussian-Wishart

  Distribution, Multivariate, Gaussian, Gaussian distribution, Multivariate gaussian distributions

Gaussian Derivatives - University at Buffalo

cedar.buffalo.edu

• Slice the surface with horizontal planes which the locus of points with the quadratic form. INFORMATION THEORY. Relative Entropy. ... great generality, and it is useful when we seek to know whether something other than the assumed case …

  Panels, Great

Multiclass Logistic Regression - University at Buffalo

cedar.buffalo.edu

Probabilistic Discriminative Models •Generative vsDiscriminative 1.Fixed basis functions in linear classification 2.Logistic Regression (two-class) 3.Iterative Reweighted Least Squares (IRLS) 4.Multiclass Logistic Regression 5.ProbitRegression 6.Canonical Link Functions 2 Machine Learning Srihari

  Model, Logistics, Regression, Logistic regression

Machine Learning Basics: Estimators, Bias and Variance

cedar.buffalo.edu

Machine Learning Basics: Estimators, Bias and Variance Sargur N. Srihari srihari@cedar.buffalo.edu ... • To distinguish estimates of parameters from their true value, a point estimate of a parameter θ is represented by • Let {x(1), x(2),..x(m)} be m independent and

  Machine, Parameters

The Hessian Matrix - University at Buffalo

cedar.buffalo.edu

Diagonal Approximation • In many case inverse of Hessian is needed • If Hessian is approximated by a diagonal matrix (i.e., off-diagonal elements are zero), its inverse is trivially computed • Complexity is O(W) rather than O(W2) for full Hessian 7

  Matrix, Diagonal, Diagonal matrix

Backpropagation - University at Buffalo

cedar.buffalo.edu

Machine Learning Srihari Matrix Multiplication: Forward Propagation •Each layer is a function of layer that preceded it •First layer is given by z =h(W(1)T x +b(1)) •Second layer is y = σ(W(2)T x +b(2)) •Note that W is a matrix rather than a vector

Machine Learning Basics: Supervised Learning Algorithms

cedar.buffalo.edu

Deep Learning Probabilistic Supervised Classification Srihari • If we only have two classes we only need to specify the distribution for one of these classes – The probability of the other class is known – Linear regression has a closed-form solution – But …

  Learning

Iterative Reweighted Least Squares

cedar.buffalo.edu

•Derivative has the form •Setting equal to zero and solving we get Machine Learning Srihari 4 y(x,w)=w j ... 3.Gradient 4.Hessian 5.Newton-Raphsonupdate …

  Derivatives