Course work

My course work reflects my scientific and research interests:

• Statistical modeling, machine learning and data mining

• Control, estimation, optimization

• Traffic theory

Statistical modeling, machine learning and data mining

Behavioral data mining (Auditing, Spring 2012, UC Berkeley)

Large-scale data mining of behavioral data - data generated by people. Examples include shopping (Amazon, Ebay), messaging (Facebook, Twitter, Livejournal), tagging (Flickr, Digg) repositories (Wikipedia, Stack Over Flow) and recommenders (NetFlix, Amazon etc). Tools include Hadoop and MarkLogic 5. Topics and algorithms include sentiment analysis, scientific computing, excavating (crawling, web services), clustering, model tuning, prediction, network algorithms (HITS and pagerank), visualization of large datasets.
Taught by John Canny

Probability for applications (Fall 2011, UC Berkeley)

Simple Random Walks, Generating Functions (Probability generating functions, Characteristic functions, Central Limit Theorem), Modes of convergence (Convergence in distribution, probability, mean-square, almost sure, strong law of large numbers), Branching Processes (Extinction, Long-time Behavior), Poisson Processes (Poisson Approximation and the Stein-Chen method, Thinning and Superposition, Campbell-Hardy Theorem), Markov Chains (Stationary distributions, Markov Chain Monte Carlo, Metropolis-Hastings, Continuous time Markov Chains), Martingales (Optional stopping, Azuma-Hoeffding inequality), Diffusions (Brownian motion, Continuum limits of birth and death chains, Ito's formula).
Taught by Allan Sly

Bayesian Modeling and Inference (Spring 2010, UC Berkeley)

Priors (conjugate, noninformative, reference; Hierarchical models, spatial models, longitudinal models, dynamic models, survival model;Testing; Model choice; Inference (importance sampling, MCMC, sequential Monte Carlo); Decision theory and frequentist perspectives (complete class theorems, consistency, empirical Bayes); Experimental design; Nonparametric models (Dirichlet processes, Gaussian processes, neutral-to-the-right processes, completely random measures)
Taught by Michael Jordan

Graphical Model theory (Fall 2009, UC Berkeley)

Classification regression, clustering, dimensionality, reduction, and density estimation. Mixture models, hierarchical models, factorial models, hidden Markov, and state space models, Markov properties, and recursive algorithms for general probabilistic inference nonparametric methods including decision trees, kernal methods, neural networks, and wavelets. Ensemble methods.
Taught by Peter Bartlett

Practical Machine Learning (Fall 2009, UC Berkeley)

Classification, regression, clustering, dimensionality reduction, feature selection, hidden markov models, graphical models, active learning, experimental design, reinforcement learning, bootstrap, cross-validation, ROC plots, time series, sequential hypothesis testing, anomaly detection, bayesian non parametric methods (Dirichlet processes), optimization methods for learning.

Random processes in systems (Fall 2009, UC Berkeley)

Probability, random variables and their convergence, random processes. Filtering of wide sense stationary processes, spectral density, Wiener and Kalman filters. Markov processes and Markov chains. Gaussian, birth and death, poisson and shot noise processes. Elementary queueing analysis. Detection of signals in Gaussian and shot noise, elementary parameter estimation.
Taught by Michael Gastpar

Statistics (Spring 2006)

Statistical modelling, estimation bounds, estimators, confidence tests and intervals, asymptotic properties.

Control, estimation, optimization

Distributed optimization (Spring 2011, UC Berkeley)

The advent of massive data sets and hugely parallel computing offers novel opportunities for distributed optimization. While much of the theory underpinning this area has been around for decades, there has been tremendous progress (and new challenges!) recently, and it is time to offer a new look at the subject. First, the class explores recent advances in first-order methods for convex optimization, which constitute the main building block for many of the more advanced algorithms developed later. Second, the class focuses on algorithms for distributed optimization under computation and communication constraints. Our starting point here is mathematical decomposition techniques traditionally developed for exploiting structure in large-scale optimization. The resulting algorithms are applicable to a cluster of collocated computers under centralized coordination. We then discuss algorithms that allow for inaccurate computations and delayed or restricted communication, as well as cooperative and non-cooperative algorithms for distributed optimization in the absence of centralized control. These algorithms are applicable to a cloud of distributed computers with partial and possibly time-varying communication, and to networks of devices found in communication systems and robotics.
Taught by Laurent El Ghaoui, Mikael Johansson and John Duchi

Introduction to Convex Optimization (Fall 2010, UC Berkeley)

Nonlinear optimization problems where the objective to be minimized, and the constraints, are both convex. Convex optimization theory and algorithms, applications arising in engineering design, machine learning and statistics, finance, and operations research. Different classes of convex optimization problems (LP, QP, SCSP, SOCP, SDP). Use of convex programming to address hard, non convex problems (such as “combinatorial optimization” problems).
Taught by Laurent El Ghaoui

Nonlinear Systems–Analysis, Stability and Control (Spring 2010, UC Berkeley)

Second Order systems, stability, Popov, circle criterion, Input-Output stability, passivity, bifurcations of dynamical systems, introduction to the differential geometry.
Taught by Murat Arcak

Linear Systems theory (Fall 2009, UC Berkeley)

Basic system concepts, state-space and I/O representation, properties of linear systems. Controllability, observability, minimality, state and output-feedback. Stability. Observers. Characteristic polynomial. Nyquist test.
Taught by Claire Tomlin

Control : Basic concepts and applications in mechanics (Spring 2007, Ecole Polytechnique)

automation, dynamic systems, controllability stabilisation and observability in linear systems; optimal control, trajectory optimization, Pontryagin's maximum principle, Hamilton-Jacobi-Bellman equation
Taught by Frederic Bonnans and Yacine Chitour

Optimal design of structures (Spring 2007, Ecole Polytechnique)

Parametric shape optimisation, geometric shape optimisation, topology optimisation by the homogenisation method, topology optimisation using evolutionary algorithms
Taught by Gregoire Allaire

Operations Research : Mathematical Aspects and Applications (Fall 2006, Ecole Polytechnique)

Modelling combinatorial optimisation problems, network flows, potential flows, Dantzig-Wolfe algorithm, interior point methods, branch and bound techniques, relaxations and cuts, lagrangian relaxation, benders decomposition, semidefinite programming, dynamic programming.
Taught by Stephane Gaubert and Frederic Bonnans

Numerical analysis and optimization (Fall 2005, Ecole Polytechnique)

Finite differences, variational formulations, Lax-Milgram lemma, Sobolev spaces, weak solutions. eigenvalues and eigenfunctions, finite element method, optimization in finite or infinite dimension, linear programming.
Taught by Gregoire Allaire and Pierre-Louis Lions

Distribution Theory, Fourier Analysis and Dynamical Systems (Fall 2005, Ecole Polytechnique)

Differential equations, linearization, perturbation theory, sub manifolds, differential forms, topological applications, characteristics method. Distributions theory, Fourier transform, Poisson and Laplace equations, harmonic functions.
Taught by Frederic Bony and Claude Viterbo.

Transportation engineering

Urban transport management (Spring 2008, University Paris Est, France)

Historical overview, the market of urban movements, organisation and funding of collective urban transport systems, economic control of the transport system and managerial challenge, urban transport systems, service quality.
Taught by Jean Laterrasse

Traffic engineering (Fall 2007, University Paris Est, France)

Microscopic and macroscopic traffic variables, fundamental diagram, statistical distributions, traffic sensors and new technologies of probe vehicles, kinematics waves and shock waves, car following theory and traffic simulation models. Traffic signal settings and Webster's method, delays at traffic signals, queuing models, urban traffic control practices: green waves and traffic signal plans, traffic management and control practices on motorway networks, automatic incident detection.
Taught by Simon Cohen

Transport demand modelling (Fall 2007, University Paris Est, France)

Selection models, transport mode selection, choice of itinerary in a network, modelling collective and plurimodal transport processes, spatial distribution of trips: statistical models, economic models, movement flow formation, generation models, planning and prospects.
Taught by Fabien Leurent

Transportation systems (Fall 2007, University Paris Est, France)

Conception and exploitation of transportation systems.
Taught by Jean Laterrasse.

Transport and sustainable development (Fall 2007, University Paris Est, France)

Economy related to environment (externality, optimal pollution), evaluation of external costs, economic and socio-political study of transportation noises, policies and regulation.
Taught by Jean-Pierre Orfeuille