Guide to Faculty Research Interests

Class of 2021 Senior Thesis and Junior Independent Work

Dr. Robert Almgren
Lecturer, ORF 474 Spring 2020
ralmgren@princeton.edu

Research Interests

Financial market microstructure, high frequency trading, high frequency data analysis, optimal trade execution, and machine learning methods for all of the above.

Professor Amirali Ahmadi
Room 329, Sherrerd Hall - Ext. 6416
aaa@princeton.edu

Research Interests

  • Optimization: algebraic methods in optimization, semidefinite programming, polynomial optimization.
  • Computational aspects of dynamics and control: Optimization-based Lyapunov theory for verification of dynamical systems.
  • Control-oriented learning: Learning dynamical systems from trajectory observations subject to side information.
  • Algorithms and complexity: Computational complexity in numerical optimization, convex relaxations in combinatorial optimization.
  • I am also interested in applications of these tools to semialgebraic problems in systems theory, machine learning, robotics, and economics.

Professor Rene Carmona (On Sabbatical 2020-21)
Room 210, Sherrerd Hall - Ext. 2310
rcarmona@princeton.edu

Research Interests

Stochastic analysis, stochastic control and stochastic games, especially mean field games. High Frequency markets, environmental finance and energy and commodity markets.

Professor Matias Cattaneo
Room 227, Sherrerd Hall - Ext. 8825
cattaneo@princeton.edu

Research Interests

Econometric theory and mathematical statistics; program evaluation and treatment effects; machine learning, nonparametric and semiparametric methods; high-dimensional inference; applications to social and behavioral sciences.

Professor Jianqing Fan
Room 205, Sherrerd Hall - Ext. 7924
jqfan@princeton.edu

Research Interests

Finance, machine learning, statistics, portfolio choices, financial risks, computational biology, among others.

Junior Independent Work/Senior Thesis Topics:

  1. Statistical data analysis
  2. Finance (Portfolio selection, asset pricing, financial risks, high-frequency trading)
  3. Machine learning (deep learning, networks, topic modeling)
  4. Biostatistics and genomics

Professor Boris Hanin (Coming to ORFE September 2020)
bhanin@princeton.edu

Research Interests

Theory of Deep Learning: Neural networks are machine learning models that have achieved state of the art in a variety of practical tasks. I am interested in elucidating the principles by which they work, understanding how to make them better, and extracting the general lessons they hold for statistics and machine learning. A key question is: at finite but large depth and width, what is the statistical behavior of neural networks at initialization? This concerns inductive bias, aligning priors to datasets, and setting hyperparameters such as architecture, learning rate, non-linearity, batch size, etc in a principled way. In addition to such probabilistic/statistical questions, I am also fascinated by understanding the roles of data augmentation and large vs. small learning rates. The questions can involve a range of tools from random matrix theory to combinatorics, the renormalization group, and high dimensional probability.

Semiclassic Analysis: Bohr's correspondence principle says that in the limit of large quantum numbers, quantum mechanics should begin to resemble classical mechanics. For example, by studying a quantum mechanical system in the limit where Planck's constant h tends to zero (the so-called semiclassical limit), the high energy behavior of quantum states should resemble the long-time behavior of the underlying classical system. I am interested in various questions that try to make such claims precise. Usually, they involve studying the zero set and/or density of states for randomized wavefunctions at high frequency/energy. The tools involved range from riemannian geometry to Gaussian processes, geometric measure theory, spectral theory for self-adjoint operators, and special functions.

Random Polynomials: The location and behavior of the zeros of polynomials are a classical topic math and science. It is a famous theorem that for polynomials of degree five and higher in one variable there is no formula for their zeros as a function of their coefficients. Nonetheless, a lot can be said about the zeros of a polynomial when it is chosen at random. For example, a result due to Gauss is that if p_n is a degree n polynomial of one complex variable, then its critical points (solutions to p_n'(z)=0) are inside the convex hull of its zeros. It turns out that much more is true with high probability if p_n is chosen at random. Namely, most of its zeros have a unique nearby critical point. Amazingly, this rather surprising fact was not known until a few years ago, and much remains to be understood! The tools here involve probability and complex analysis.

Dr. Margaret Holen
Lecturer, ORF 473 Spring 2020
holen@princeton.edu

Research Interests

As a finance industry practitioner and Goldman Sachs partner, I have extensive project supervision experience related to financial markets and investment banking applications. My current industry focus is on advising and investing in early-stage FinTech companies, with an emphasis on “alternative” data and machine learning. In addition, I am currently teaching an ORFE course focused on FinTech in consumer lending, which explores how evolving data sets and ML toolkits are changing that sector, and I have interest in a variety of related research topics.

Current senior thesis topics under supervision (’18-’19) include:

  • Information Cascades and Herding in Online Restaurant Reviews, with work spanning data analysis (of Yelp reviews), simulation and mathematical theory. (Motivated by the student’s active interest in online review systems, as an avid user.)
  • Industry Surveys of Technology Spending: Extracting Information on Revenue Projection applies stats/ML methods to a proprietary dataset and includes theory related to wisdom of crowds and detecting expert from crowds. (Motivated by the student’s summer internship work.)
  • Information from Earnings Calls: Applying natural language processing methods from earnings calls to understand information content relative to other financial and markets data. (Motivated by a popular article that cited a variety of related studies.) Previous senior thesis topics under supervision (’18-’19):
  • Machine Learning methods applied to employee GlassDoor reviews. (Motivated by the student’s interest in start-ups and how employee satisfaction can help VC’s direct capital. Supervisor helped source data from an alternative data startup.)
  • Who gets the money? Analysis of ETF fund sizes relative to fund attributes and performance. (Motivated by student’s curiosity about the sector, the specific topic originated with an advisor suggestion.)
  • Team choice for competitive electronic gaming using ML methods. (Motivated and shaped by the student’s hobby interests.)

Professor Jason Klusowski (Coming to ORFE September 2020)
jason.klusowski@princeton.edu

Research Interests

I am broadly interested in the theory and application of various statistical learning models, in particular tree-structured models (e.g., random forests, gradient tree boosting, CART), neural networks, statistical network models, and latent variable models.

Professor Alain L. Kornhauser
Room 229, Sherrerd Hall - Ext. 4657
alaink@princeton.edu

Research Interests

Development and application of operations research and other analytical techniques in various aspects of Autonomous Vehicles, aka "SmartDrivingCars", including:

  • the fundamental design of computer vision techniques for the rapid classification and identification of the driving environment, especially “deep learning convolutional neural networks”,
  • analysis and classification of collision-free driving scenarios,
  • quantification of accident risk and the investigation, formulation and design of "pay-as-you-drive, pay-as-the-car-drives" insurance,
  • Investigation and creative design of the human-computer interfaces for SmartDrivingCars
  • operational and feasibility analyses of autonomousTaxi (aTaxi) systems

Junior Independent Work/Senior Thesis Topics: Any of the above.

Professor William A. Massey (On Sabbatical 2020-21)
Room 206, Sherrerd Hall - Ext. 7384
wmassey@princeton.edu

Research Interests

The applications of and theory for the dynamics of resource sharing. Motivating examples include the design and management of communication networks as well as healthcare systems. Research interests are in dynamic rate queueing theory, stochastic networks, dynamical systems, optimal control, Monte-Carlo simulation, and time-inhomogeneous Markov processes.

Professor John M. Mulvey
Room 207 Sherrerd Hall – Ext. 5423
mulvey@princeton.edu

Research Interests

Large-scale stochastic optimization models, algorithms, and applications, especially financial planning and wealth management. Multi-period financial planning applications for large insurance companies, hedge funds, global FinTech firms, and individuals. Apply novel methods in machine learning to financial planning systems. Combining advanced mathematical financial models with deep neural networks to address transaction costs and overcome the exponential growth in model size.

Conducting research with Ant Financial (Alibaba) on enterprise risk management for a global FinTech firm, and several large banks in San Francisco and New York City, and a multi-manager hedge fund in Austin Texas.

Junior Independent Work/Senior Thesis Topics:

  1. Stochastic optimization problems under uncertainty in finance. Multi-stage portfolio models for stocks, index fund generation, and international currency. Achieving wide diversification via advanced policy rules. Solving the resulting massive stochastic optimization model with advanced algorithms in nonlinear programming and machine learning.
  2. Modeling alternative asset categories. Over the past two decades, institutional investors have shifted capital to the so-called alternative asset categories, including private equity, hedge funds, leveraged loans, real assets, and others. These securities are generally hybrid in nature, with several embedded risk factors. We explore these topics, employing advanced topics in machine learning and stochastic optimization.
  3. Model comparison. All too often, operations researchers and decision analysts believe that their models are pure representations of the “scientific method.” Yet two modelers will likely employ two different models with varying recommendations when confronted with a decision problem. These differences are often due to behavioral bias and value considerations. Discover a real-world application and study its ramifications.
  4. Educational aspects of financial planning. Most individuals are woefully under prepared for making significant financial decisions. There are many behavioral biases and unfortunate habits that lead to under performance. These issues are becoming more significant with the rise in automated financial planning systems (robo-advisors).

Professor Miklos Racz
Room 204, Sherrerd Hall - Ext. 8281
mracz@princeton.edu

Research Interests

My research focuses on probability, statistics, and its applications. I am interested in statistical inference problems in complex systems, in particular on random graphs and in genomics. I am also interested more broadly in applied probability, combinatorial statistics, information theory, control theory, interacting particle systems, and voting.

Professor Mykhaylo Shkolnikov
Room 202, Sherrerd Hall - Ext. 1044
mykhaylo@princeton.edu

Research Interests

  1. Stochastic portfolio theory
  2. Optimal Investment
  3. Stochastic analysis
  4. Interacting particle systems
  5. Random matrix theory
  6. Large deviations
  7. Markov chains

Professor Ronnie Sircar
Room 208, Sherrerd Hall - Ext. 2841
sircar@princeton.edu

Research Interests

Financial Mathematics & Engineering; stochastic models, especially for market volatility; optimal investment and hedging strategies; analysis of financial data; credit risk; dynamic game theory and oligopoly models; energy and commodities markets; reliability of the electricity grid under increased use of solar and wind technologies.

Professor Mete Soner
Room 225, Sherrerd Hall, Ext.- 5130
soner@princeton.edu

Research Interests

Mathematical theory of optimal control and decisions under uncertainty, and applications of stochastic optimization techniques in economics, financial economics and quantitative finance, and high-dimensional computational problems. Current topics are:

  1. Numerical methods for high-dimensional Markov Decision problems.
  2. McKean-Vlasov optimal control problems.
  3. Model-independent and robust finance. In particular, pricing and hedging of complex financial instruments using market data with little or no model assumptions.
  4. Theoretical foundations of financial markets with Knightian uncertainty.
  5. Hedging in markets with frictions such as illiquidity or transaction costs.

Professor Bartolomeo Stellato (Coming to ORFE September 2020)
bstellato@princeton.edu

Research Interests

Development of theory and data-driven computational tools for mathematical optimization, machine learning and optimal control. Applications include control of fast dynamical systems, finance, robotics and autonomous vehicles. Mathematical optimization - Large-scale and embedded convex optimization - First-order methods for sparse, low-rank and combinatorial optimization - Differentiable optimization Machine learning for optimization - Learning heuristics to accelerate combinatorial optimization algorithms - Learning solutions of optimization problems with varying data Control systems - Learning control policies for continuous and hybrid systems - High-speed online optimization for real-time control

Professor Ludovic Tangpi
Room 203 Sherrerd Hall – Ext. 4558
ludovic.tangpi@princeton.edu

Research Interests

My research interests are in stochastic calculus and mathematical finance. More specifically, my work in stochastic analysis aims at developing the understanding of stochastic control and stochastic differential games using probabilistic arguments. Such control and differential games arise in a variety of optimal decision applications, including optimal investment, economics, engineering or biology. Some of the main issues here concern the analysis of optimal decision policies in diffusion models, the study of representations that enable their efficient numerical computations, and similar questions for games with a large number of players.

In mathematical finance, I focus on developing the theory of quantitative risk management. Here, I put a particular emphasis on computational aspects, and develop tools needed to achieve efficient computation. In fact, financial agents (especially banks and insurance companies) are eager to evaluate the riskiness of their decision and often evaluate associated risks periodically. Therefore, it is essential to understand computational issues: When using a particular model, how to select it? Can we use the specificities of a particular model to improve the estimation? How do we account for model uncertainty? It is possible to build model-free methods? Or statistical estimations entirely data-driven? Can we derive theoretical guarantees and convergence rates? This line of work includes (and is not limited to) elements of probability theory, stochastic calculus, as well a numerical and data analysis.

Professor Robert J. Vanderbei
Room 209, Sherrerd Hall - Ext. 0876
rvdb@princeton.edu

Research Interests

Large-scale mathematical optimization models arising in engineering and corresponding solution methodologies.

Junior Independent Work/Senior Thesis Topics

  1. Trajectory optimization problems to find orbits in celestial mechanics.
  2. Fast Fourier Optimization with applications in optics, acoustics, digital filtering, etc.
  3. Parametric simplex method with applications in image analysis.

Professor Ramon van Handel
Room 207, Fine Hall – Ext. 3791
rvan@princeton.edu

Research Interests

I am broadly interested in probability theory and its interactions with other fields. Probability theory, i.e., the study of randomness, is a very rich subject: it combines many different types of mathematics, and is used to solve a surprisingly diverse range of problems in different fields. I am particularly fascinated by the development of probabilistic principles and methods that explain the common structure in a variety of pure and applied mathematical problems.