{"id":726,"date":"2020-02-22T13:04:11","date_gmt":"2020-02-22T10:04:11","guid":{"rendered":"http:\/\/users.utu.fi\/ionpet\/?page_id=726"},"modified":"2020-04-24T12:49:18","modified_gmt":"2020-04-24T09:49:18","slug":"foundations-of-machine-learning","status":"publish","type":"page","link":"https:\/\/users.utu.fi\/ionpet\/foundations-of-machine-learning\/","title":{"rendered":"Foundations of Machine Learning"},"content":{"rendered":"<h3>Course name: MATE5424-3001 Foundations of Machine Learning<\/h3>\n<h4>Short description<\/h4>\n<p>This course focuses on the mathematical foundations of basic machine learning concepts and algorithms. Using mathematical language we aim to express widely used machine learning concepts that seem intuitively obvious, but turn out to be surprisingly difficult to use optimally in practice. The aim is to gain insights into several basic machine learning tasks, to understand what they do, what they are best at, and what their limitations are.<\/p>\n<p>The course is an excellent introduction to machine learning for mathematics students. It is also highly suitable to computer science students as a companion to the machine learning engineering courses, providing the mathematical background to the algorithmics and the programming methods they introduce.<\/p>\n<h4>Course textbooks<\/h4>\n<ul>\n<li>Marc Peter Deisenroth, A. Aldo Faisal, and Cheng Soon Ong. <a href=\"https:\/\/mml-book.com\/\">Mathematics for Machine Learning<\/a>. Cambridge University Press, 2020.<\/li>\n<li>Mehryar Mohri, Afshin Rostamizadeh, and Ameet Talwalkar. <a href=\"https:\/\/cs.nyu.edu\/~mohri\/mlbook\/\">Foundations of Machine Learning<\/a>. MIT Press, Second Edition, 2018.<\/li>\n<li>\n<div class=\"page\" title=\"Page 1\">\n<div class=\"layoutArea\">\n<div class=\"column\">\n<p>Avrim Blum, John Hopcroft, and Ravindran Kannan. <a href=\"https:\/\/www.cs.cornell.edu\/jeh\/book.pdf\">Foundations of Data Science<\/a>. Cambridge University Press, 2020.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/li>\n<\/ul>\n<h4>Mathematical background<\/h4>\n<p>The mathematical background needed in the course is well presented in the first part (&#8220;Mathematical Foundations&#8221;) of the Deisenroth et al <a href=\"https:\/\/mml-book.com\/\">textbook<\/a>. It is also well presented in the appendices A-E of the Mohri et al <a href=\"https:\/\/cs.nyu.edu\/~mohri\/mlbook\/\">textbook<\/a>. We assume the students will consult these books whenever needed. Only essential parts of the mathematical background will be introduced in the lectures.<\/p>\n<h4>Syllabus<\/h4>\n<ul>\n<li>Supervised learning\n<ul>\n<li>Linear regression<\/li>\n<li>Generalised linear regression (incl. logistic regression)<\/li>\n<li>Classification with support-vector machines<\/li>\n<\/ul>\n<\/li>\n<li>Unsupervised learning\n<ul>\n<li>Density estimation (incl. clustering)<\/li>\n<\/ul>\n<\/li>\n<li>Dimensionality reduction\n<ul>\n<li>Principal component analysis<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<h4>Lectures<\/h4>\n<ul>\n<li><a href=\"https:\/\/users.utu.fi\/ionpet\/wp-content\/uploads\/sites\/931\/2020\/03\/FoundML_Lecture_1_v2020.pdf\">Introduction<\/a><\/li>\n<li>8. Models and data<\/li>\n<li>9. Linear regression\n<ul>\n<li>9.1 Problem formulation<\/li>\n<li>9.2 Parameter estimation\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/b0228c7c1b47492abbfc\/\">9.2.2 Overfitting in linear regression<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/dc6f2ae1660f44bf8f4a\/\">9.2.3 Maximum a posteriori estimation<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/a979fef3acc449aa9b7f\/\">9.2.4 Regularised least squares. LASSO<\/a><\/li>\n<\/ul>\n<\/li>\n<li>9.3 Bayesian linear regression\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/8c3b81482c304ffeaa38\/\">9.3.1 Model<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/d2f612fdab1841ebb27f\/\">9.3.2 Prior predictions<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/806cffe889c24355b597\/\">9.3.3 Posterior distribution<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/6f466c98015847b990d6\/\">9.3.4 Posterior predictions<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/0cfd6bfbb80a446abe5a\/\">9.3.5 Marginal likelihood<\/a><\/li>\n<\/ul>\n<\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/15d84e786c814e7fa010\/\">9.4 A brief geometric view on linear regression<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/5af2182805274de68523\/\">9.5 Generalised linear models<\/a> (e.g., logistic regression and neural networks)<\/li>\n<\/ul>\n<\/li>\n<li>10. Dimensionality reduction with principal component analysis\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/a8d814a54e4643b09009\/\">10.1 Problem setting<\/a><\/li>\n<li>10.2 PCA as a problem of maximising the data variance\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/f48b835e174e447ba45c\/\">Finding the direction with the single highest variance<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/e723c1a902a547e38360\/\">M-dimensional subspace with maximal variance<\/a><\/li>\n<\/ul>\n<\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/fd76b628865b4095b53e\/\">10.3 PCA as a problem of minimising the average reconstruction error<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/1ce0b0acd5c244cfa464\/\">10.5 PCA in high dimensions<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/696a848b91e446ba98ed\/\">10.6 Key steps of PCA in practice<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/d209c9dac32f4c098ec9\/\">10.7 Probabilistic PCA<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/b0d5127f36ad4993b8b5\/\">10.8 Connections to other topics<\/a><\/li>\n<\/ul>\n<\/li>\n<li>11. Density estimation with Gaussian mixture models\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/bbd67eff1bdb41938b62\/\">11.0 Introduction<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/d8122ae9108a4c68883c\/\">11.1 Model formulation<\/a><\/li>\n<li>11.2 Parameter learning with maximum likelihood\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/93e12ec00cb345a9a938\/\">11.2.1 Responsibilities<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/4104125fab4e43cf95a4\/\">11.2.2 Updating the means<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/73f7539b41bc411a966e\/\">11.2.3 Updating the covariances<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/66c0ace4207c4d03ab41\/\">11.2.4 Updating the weights<\/a><\/li>\n<\/ul>\n<\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/0c65adbb60b4425c8e8d\/\">11.3 The EM algorithm<\/a><\/li>\n<li>11.4 The latent-variable perspective on GMM\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/3793f1d77f15437d9a3d\/\">11.4.1 The probabilistic model<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/d1ebc72b6c98421dac06\/\">11.4.2 The likelihood<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/e95ff02536394961a659\/\">11.4.3 &#8211; 11.4.4 The posterior distribution<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/d26db9c7494943838864\/\">11.4.5 The EM algorithm revisited<\/a><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<\/li>\n<li>12. Classification with support vector machines\n<ul>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/782e8e0f3ab4418baf62\/\">12.1 Separating hyperplanes<\/a><\/li>\n<li>12.2 Hard-margin SVM: <a href=\"https:\/\/seafile.utu.fi\/f\/796f9b8d55964581a3a3\/\">part 1<\/a>, <a href=\"https:\/\/seafile.utu.fi\/f\/ea2d2c22845541b3a49d\/\">part 2<\/a><\/li>\n<li>12.2 Soft-margin SVM: <a href=\"https:\/\/seafile.utu.fi\/f\/b3601cddfc474b36b251\/\">geometric view<\/a>, <a href=\"https:\/\/seafile.utu.fi\/f\/d269c7877bb948fbb0fe\/\">loss function view<\/a><\/li>\n<li>12.3 Dual SVM: <a href=\"https:\/\/seafile.utu.fi\/f\/ab7cd8c5e561421cb184\/\">via Lagrange multipliers<\/a>, <a href=\"https:\/\/seafile.utu.fi\/f\/680abd325a544d18b02a\/\">via convex hulls<\/a><\/li>\n<li><a href=\"https:\/\/seafile.utu.fi\/f\/7de1f72e5db947feac79\/\">12.4 Kernels<\/a><\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<h4><\/h4>\n<h4>Course feedback<\/h4>\n<p>The Department of Mathematics and Statistics is collecting periodic feedback on its courses. It is important for the department and for the lecturer to get this feedback. The questionnaire will only take a few minutes to answer. The questions (in Finnish) can be found at <a href=\"https:\/\/webropol.com\/s\/mattilpalauteIV2020\">https:\/\/webropol.com\/s\/mattilpalauteIV2020<\/a><\/p>\n<p>&nbsp;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Course name: MATE5424-3001 Foundations of Machine Learning Short description This course focuses on the mathematical foundations of basic machine learning concepts and algorithms. Using mathematical&hellip;<\/p>\n","protected":false},"author":4292,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-726","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/pages\/726","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/users\/4292"}],"replies":[{"embeddable":true,"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/comments?post=726"}],"version-history":[{"count":52,"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/pages\/726\/revisions"}],"predecessor-version":[{"id":891,"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/pages\/726\/revisions\/891"}],"wp:attachment":[{"href":"https:\/\/users.utu.fi\/ionpet\/wp-json\/wp\/v2\/media?parent=726"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}