- You are here:
- GT Home
- Home
- News & Events

Series: ACO Student Seminar

In network routing users often tradeoff different objectives in selecting their best route. An example is transportation networks, where due to uncertainty of travel times, drivers may tradeoff the average travel time versus the variance of a route. Or they might tradeoff time and cost, such as the cost paid in tolls.

We wish to understand the effect of two conflicting criteria in route selection, by studying the resulting traffic assignment (equilibrium) in the network. We investigate two perspectives of this topic: (1) How does the equilibrium cost of a risk-averse population compare to that of a risk-neutral population? (i.e., how much longer do we spend in traffic due to being risk-averse) (2) How does the equilibrium cost of a heterogeneous population compare to that of a comparable homogeneous user population?

We provide characterizations to both questions above.

Based on joint work with Richard Cole, Thanasis Lianeas and Nicolas Stier-Moses.

At the end I will mention current work of my research group on algorithms and mechanism design for power systems.

**Biography: ** Evdokia Nikolova is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Texas at Austin, where she is a member of the Wireless Networking & Communications Group. Previously she was an Assistant Professor in Computer Science and Engineering at Texas A&M University. She graduated with a BA in Applied Mathematics with Economics from Harvard University, MS in Mathematics from Cambridge University, U.K. and Ph.D. in Computer Science from MIT.

Nikolova's research aims to improve the design and efficiency of complex systems (such as networks and electronic markets), by integrating stochastic, dynamic and economic analysis. Her recent work examines how human risk aversion transforms traditional computational models and solutions. One of her algorithms has been adapted in the MIT CarTel project for traffic-aware routing. She currently focuses on developing algorithms for risk mitigation in networks, with applications to transportation and energy. She is a recipient of an NSF CAREER award and a Google Faculty Research Award. Her research group has been recognized with a best student paper award and a best paper award runner-up. She currently serves on the editorial board of the journal Mathematics of Operations Research.

Series: ACO Student Seminar

In an optimal design problem, we are given a set of linear experiments v1,...,vn \in R^d and k >= d, and our goal is to select a set or a multiset S subseteq [n] of size k such that Phi((\sum_{i \in [n]} v_i v_i^T )^{-1}) is minimized. When Phi(M) = det(M)^{1/d}, the problem is known as the D-optimal design problem, and when Phi(M) = tr(M), it is known as the A-optimal design problem. One of the most common heuristics used in practice to solve these problems is the local search heuristic, also known as the Fedorov's exchange method. This is due to its simplicity and its empirical performance. However, despite its wide usage no theoretical bound has been proven for this algorithm. In this paper, we bridge this gap and prove approximation guarantees for the local search algorithms for D-optimal design and A-optimal design problems. We show that the local search algorithms are asymptotically optimal when $\frac{k}{d}$ is large. In addition to this, we also prove similar approximation guarantees for the greedy algorithms for D-optimal design and A-optimal design problems when k/d is large.

Series: ACO Student Seminar

Correlation Clustering is an elegant model that captures fundamental graph cut problems such as Minimum s-t Cut, Multiway Cut, and Multicut, extensively studied in combinatorial optimization.

Here, we are given a graph with edges labeled + or - and the goal is to produce a clustering that agrees with the labels as much as possible: + edges within clusters and - edges across clusters.

The classical approach towards Correlation Clustering (and other graph cut problems) is to optimize a global objective, e.g., minimizing the total number of disagreements or maximizing the total number of agreements.

We depart from this and study local objectives: minimizing the maximum number of disagreements for edges incident on a single node, and the analogous max min agreements objective.

This naturally gives rise to a family of basic min-max graph cut problems.

A prototypical representative is Min-Max s-t Cut: find an s-t cut minimizing the largest number of cut edges incident on any node.

In this talk we will give a short introduction of Correlation Clustering and discuss the following results:

- an O(\sqrt{n})-approximation for the problem of minimizing the maximum total weight of disagreement edges incident on any node (thus providing the first known approximation for the above family of min-max graph cut problems)
- a remarkably simple 7-approximation for minimizing local disagreements in complete graphs (improving upon the previous best known approximation of 48)
- a (1/(2+epsilon))-approximation for maximizing the minimum total weight of agreement edges incident on any node, hence improving upon the (1/(4+epsilon))-approximation that follows from the study of approximate pure Nash equilibria in cut and party affiliation games.

Joint work with Moses Charikar and Neha Gupta.

Series: ACO Student Seminar

The popularity of machine learning is increasingly growing in transportation, with applications ranging from traffic engineering to travel demand forecasting and pavement material modeling, to name just a few. Researchers often find that machine learning achieves higher predictive accuracy compared to traditional methods. However, many machine-learning methods are often viewed as “black-box” models, lacking interpretability for decision making. As a result, increased attention is being devoted to the interpretability of machine-learning results.

In this talk, I introduce the application of machine learning to study travel behavior, covering both mode prediction and behavioral interpretation. I first discuss the key differences between machine learning and logit models in modeling travel mode choice, focusing on model development, evaluation, and interpretation. Next, I apply the existing machine-learning interpretation tools and also propose two new model-agnostic interpretation tools to examine behavioral heterogeneity. Lastly, I show the potential of using machine learning as an exploratory tool to tune the utility functions of logit models.

I illustrate these ideas by examining stated-preference travel survey data for a new mobility-on-demand transit system that integrates fixed-route buses and on-demand shuttles. The results show that the best-performing machine-learning classifier results in higher predictive accuracy than logit models as well as comparable behavioral outputs. In addition, results obtained from model-agnostic interpretation tools show that certain machine-learning models (e.g. boosting trees) can readily account for individual heterogeneity and generate valuable behavioral insights on different population segments. Moreover, I show that interpretable machine learning can be applied to tune the utility functions of logit models (e.g. specifying nonlinearities) and to enhance their model performance. In turn, these findings can be used to inform the design of new mobility services and transportation policies.

Series: ACO Student Seminar

(The talk will be at 1-2pm, then it follows by a discussion session from 2 pm to 2:45 pm.)

Powerful AI systems, which are driven by machine learning, are increasingly controlling various aspects of modern society: from social interactions (e.g., Facebook, Twitter, Google, YouTube), economics (e.g., Uber, Airbnb, Banking), learning (e.g., Wikipedia, MOOCs), governance (Judgements, Policing, Voting), to autonomous vehicles and weapons. These systems have a tremendous potential to change our lives for the better, but, via the ability to mimic and nudge human behavior, they also have the potential to be discriminatory, reinforce societal prejudices, and polarize opinions. Moreover, recent studies have demonstrated that these systems can be quite brittle and generally lack the required robustness to be deployed in various civil/military situations. The reason being that considerations such as fairness, robustness, stability, explainability, accountability etc. have largely been an afterthought in the development of AI systems. In this talk, I will discuss the opportunities that lie ahead in a principled and thoughtful development of AI systems.

#### Bio

Nisheeth Vishnoi is a Professor of Computer Science at Yale University. He received a B.Tech in Computer Science and Engineering from IIT Bombay in 1999 and a Ph.D. in Algorithms, Combinatorics and Optimization from Georgia Tech in 2004. His research spans several areas of theoretical computer science: from approximability of NP-hard problems, to combinatorial, convex and non-convex optimization, to tackling algorithmic questions involving dynamical systems, stochastic processes and polynomials. He is also broadly interested in understanding and addressing some of the key questions that arise in nature and society from the viewpoint of theoretical computer science. Here, his current focus is on natural algorithms, emergence of intelligence, and questions at the interface of AI, ethics, and society. He was the recipient of the Best Paper Award at FOCS in 2005, the IBM Research Pat Goldberg Memorial Award in 2006, the Indian National Science Academy Young Scientist Award in 2011, and the IIT Bombay Young Alumni Achievers Award in 2016.

Series: ACO Student Seminar

We present a new general and simple method for rounding semi-definite programs, based on Brownian motion. Our approach is inspired byrecent results in algorithmic discrepancy theory. We develop and present toolsfor analyzing our new rounding algorithms, utilizing mathematical machineryfrom the theory of Brownian motion, complex analysis, and partial differentialequations. We will present our method to several classical problems, including Max-Cut, Max-di-cut and Max-2-SAT, and derive new algorithms that are competitive with the best known results. In particular, we show that the basic algorithm achieves 0.861-approximation for Max-cut and a natural variant of the algorithm achieve 0.878-approximation, matching the famous Goemans-Williamson algorithm upto first three decimal digits. This is joint work with Abbas-Zadeh, Nikhil Bansal, Guru Guruganesh, Sasho Nikolov and Roy Schwartz.

Series: ACO Student Seminar

In current convex optimization literature, there are significant gaps between algorithms that produce high accuracy (1+1/poly(n))-approximate solutions vs. algorithms that produce O(1)-approximate solutions for symmetrized special cases. This gap is reflected in the differences between interior point methods vs. (accelerated) gradient descent for regression problems, and between exact vs. approximate undirected max-flow. In this talk, I will discuss generalizations of a fundamental building block in numerical analysis, preconditioned iterative methods, to convex functions that include p-norms. This leads to algorithms that converge to high accuracy solutions by crudely solving a sequence of symmetric residual problems. I will then briefly describe several recent and ongoing projects, including p-norm regression using m^{1/3} linear system solves, p-norm flow in undirected unweighted graphs in almost-linear time, and further improvements to the dependence on p in the runtime.

Series: ACO Student Seminar

In this talk we introduce two different random graph models that produce

sparse graphs with overlapping community structure and discuss

community detection in each context. The Random Overlapping Community

(ROC) model produces a sparse graph by constructing many Erdos Renyi

random graphs (communities) on small randomly selected subsets of

vertices. By varying the size and density of these communities, ROC

graphs can be tuned to exhibit a wide range normalized of closed walk

count vectors, including those of hypercubes. This is joint work with

Santosh Vempala. In the second half of the talk, we introduce the

Community Configuration Model (CCM), a variant of the configuration

model in which half-edges are assigned colors and pair according to a

matching rule on the colors. The model is a generalization of models in

the statistical physics literature and is a natural finite analog for

classes of graphexes. We describe a hypothesis testing algorithm that

determines whether a graph came from a community configuration model or a

traditional configuration model. This is joint work with Christian

Borgs, Jennifer Chayes, Souvik Dhara, and Subhabrata Sen.

Series: ACO Student Seminar

Series: ACO Student Seminar

Consider a linear combination of independent identically distributed random variables $X_1, . . . , X_n$ with fixed weights $a_1, . . . a_n$. If the random variablesare continuous, the sum is almost surely non-zero. However, for discrete random variables an exact cancelation may occur with a positive probability. Thisprobability depends on the arithmetic nature of the sequence $a_1, . . . a_n$. We will discuss how to measure the relevant arithmetic properties and how to evaluate the probability of the exact and approximate cancelation.