Mumbai University · Final Year Computer Engineering · Rev-2019 'C' Scheme
Optimization in Machine Learning
CSDC8021 · Department Optional Course 6 · Semester VIII · 3 credits · Internal tests 20 (average of two) + end-semester 80
End-semester paper (3 hours): Six questions of 20 marks; you answer four. Each module's share of the paper follows its lecture hours.
One of the Department Optional Course 6 options.
Module 1: Background to Optimisation
4 hoursIngredients and classes of optimisation problems; types of optima; unconstrained and constrained optimisation; basics of convex optimisation.
Examined in the 80-mark end-semester paper (about 10% of it, by lecture hours); the class tests cover what has been taught by then
- Classification of optimisation problems
- Types of optima
Module 2: Derivative-based Optimisation
10 hoursUnivariate, bivariate and multivariate optimisation; convex objectives; gradient descent, conjugate gradient, momentum, Nesterov, AdaGrad and RMSProp; learning-rate schedules; Newton's method.
Examined in the 80-mark end-semester paper (about 26% of it, by lecture hours); the class tests cover what has been taught by then
- Convex objective functions
- Gradient descent and conjugate gradient
- Momentum and adaptive methods
- Newton's method
Module 3: Stochastic Methods
6 hoursNoisy descent; mesh adaptive direct search; the cross-entropy method; natural evolution strategies; covariance matrix adaptation.
Examined in the 80-mark end-semester paper (about 15% of it, by lecture hours); the class tests cover what has been taught by then
- Noisy descent and direct search
- Cross-entropy method
- Evolution strategies and CMA-ES
Module 4: Convex Optimisation
6 hoursConvex, linear and quadratic optimisation problems; geometric programming; generalised inequalities and vector optimisation.
Examined in the 80-mark end-semester paper (about 15% of it, by lecture hours); the class tests cover what has been taught by then
- Convex optimisation problems
- Linear and quadratic programming
- Geometric programming
Module 5: Evolutionary Methods
8 hoursThe generic evolutionary algorithm; chromosomes, populations and fitness; random, proportional, tournament and rank selection; elitism; stopping conditions; the canonical GA, crossover and mutation.
Examined in the 80-mark end-semester paper (about 21% of it, by lecture hours); the class tests cover what has been taught by then
- Evolutionary algorithms
- Selection operators
- Canonical genetic algorithm
Module 6: Advanced Evolutionary Methods
5 hoursGlobal-best and local-best PSO, velocity components and social structures; ant colony optimisation, stigmergy, Ant System and Ant Colony System.
Examined in the 80-mark end-semester paper (about 13% of it, by lecture hours); the class tests cover what has been taught by then
- Particle swarm optimisation
- Ant colony optimisation
Module contents are summarised from the official Mumbai University syllabus . AiCanCode.org is not affiliated with the University of Mumbai.