Lectures 2018
- Lecture September 24, 2018 - Introduction to the course. Questionnaire for checking background. Generalities on AutomaticLearning. (Ref. slide 1st-lecture)
- Lecture September 25, 2018 - The learning process. Data, learning machine. Minimization of the expected risk. Over and under fitting. (Ref. slide 2nd lecture, Bishop – Pattern Recognition and Machine Learning, Springer, 2006 (Chap 1), V. Cherlassky, F. Mulier - Learning from Data, John Wiley and Sons, 2007 (chap.1 and 2) , Bottou, Curtis, and Nocedal - Optimization Methods for LargeScale Machine Learning (section 1 and 2), Teaching Notes chapt. 1)
- Lecture October 1, 2018 - Hint on Vapnik Chervonenkis theory and Structural risk minimization principle. VC dimension of oriented hyperplanes (theorem and proof) (Ref. slide 3rd lecture, Teaching Notes chapt. 1)
- Lecture October 2, 2018 - Hyperplanes with margin - (Ref. Slide 4th lecture, Teching notes chapt.1, P. Domingos - A few useful things to know about Machine Learning). Brief review on optimization problems: basic defintions.
- Lecture October 8, 2018 - The perceptron algorithm. Example: the logical OR. (Ref. Slide 5th lecture, Teching notes chapt.2). Presentation of the Brdgestone Hackathon (deadline December 31th, 2018)
- Lecture October 9, 2018 - The perceptron algorithm (proof). The voting and the average perceptron (Ref. Slide 6th lecture, (Authors' faces), Teching notes chapt.2). Project (not mandatory) on the perceptron.
- Lecture October 15, 2018 - Beyond perceptron. Feedforward Neural Network: Multiplayer perceptron (Ref. Slide 7th-8th Lecture, Palagi - Global optimization issues in deep network regression: an overview -Section 4, Teaching notes chapt. 3)
- Lecture October 16, 2018 - Hyperparameters and weights problems (Ref. Slide 7th-8th Lecture). Python + libraries (Numpy, SkLearn, pandas, scipy.optimize) by Ruggiero Seccia (Ref. Pitone1)
- Lecture October 22, 2018 - Deep Network - Brief review of optimization algorithms. Stopping criterion
- Lecture October 24, 2018 - Backpropagation rule (slide 10th-11th Lecture)
- Lecture November 5, 2018 - Backpropagation rule - Convergence of the batch BP (learning rate) - Momentum term (Ref. Slide of 10-11th Lecture, teaching notes chapt. 3)
- Lecture November 6, 2018 - Online methods for MLP: incremental, stochastic gradient, batch method (Ref. Slide of the 12th lecture; in depth reading: Optimization methods for large-scale machine learning by L. Bottou, FE Curtis, J Nocedal)
- Lecture November 9, 2018 - Decomposition methods for MLP: Extreme learning (Ref. Slide of the 13th lecture)
- Lecture November 12, 2018 - Decomposition methods for MLP (Ref. Slide of the 14th lecture)
- Lecture November 13, 2018 - Regularized RBF; generalized RBF network; learning paradigm for RBF network: unsupervised versus supervised selection of the centers (Ref. Slide of the 15th lecture; Teaching notes chapt. 4)
- Lecture November 13, 2018 - Supervised selection of the centers Full optimization. Two block decompostion methods, exact and inexact solution of subproblems, convergence properties. Decomposition methods: block learning of centers (Ref. Slide of the 16-17 th lecture; Teaching notes chapt. 4)
- Lecture November 16, 2018 - Two block decompostion methods, exact and inexact solution of subproblems, convergence properties. Decomposition methods: block learning of centers (Ref. Slide of the 16-17th lecture; Teaching notes chapt. 4)
- MIDTERM November 20, 2018
- Lecture November 23 2018 - Hard SVM: generalities, defintion of margin, defintion of the max margin problem (Ref. Slide 19th lecture). The primal hard SVM problem
- Lecture November 26 2018 - Convex optimization: KKT conditions for linearly constrained porblems. Feasible and descent directions. Frank-wolfe conditional gradient
- Lecture November 27 2018 - Duality in convex quadratic programming. The Wolfe dual. The weak duality theorem. Construction of the dual Hard SVM problem
- Lecture December 3, 2018 - Soft SVM. The dual problem of C-SVM
- Lecture December 4, 2018 -
- Lecture December 10, 2018 - Dual problem optimality conditions (KKT and feasible&descent directions)
- Lecture December 11, 2018 - Definition of kernel (non linear SVM). Analytic solution of QP in two dimension. Use of Quadrprogr. Tutorial on TensorFlow and Keras (Scipy, class1, class2)
- Lecture December 14, 2018
- Lecture December 17, 2018
- Lecture December 18, 2018 - Final Term