Department Seminar

Speaker : Dr. Arindam Khan

Researcher, IDSIA SUPSI

Switzerland

Title : Approximating Geometric Knapsack via L-packings

Date : Monday, August 21, 2017

Time : 11:00 AM

Venue : CSA Seminar Hall (Room No. 254, First Floor)

Live Video : MMCRs - ECE, CEDT, EE

or

URL -- http://mmcr.iisc.ernet.in:8008/live/csa.html

Abstract

We study the two-dimensional geometric knapsack problem (2DK), a geometric

variant of the classical knapsack problem. In this problem, we are given a

set of axis-aligned rectangular items, each one with an associated profit,

and an axis-aligned square knapsack. The goal is to find a (non-overlapping)

packing of a maximum profit subset of items inside the knapsack without

rotating items. This is a very well-studied optimization problem and finds

applications in scheduling, memory allocation, advertisement placement etc.

The best-known polynomial-time approximation factor for this problem (even just

in the cardinality case) is $2+eps$ [Jansen and Zhang, SODA 2004].

After more than a decade, in this paper we break the 2-approximation barrier,

achieving a polynomial-time $17/9+eps<1.89$ approximation, which improves to

$558/325+eps<1.72$ in the cardinality case.

We also consider the variant of the problem with rotations (2DKR) , where the

items can be rotated by $90$ degrees. Also in this case the best-known polynomial-time

approximation factor (even for the cardinality case) is $2+eps$ [Jansen and Zhang,

SODA 2004]. Exploiting part of the machinery developed for 2DK plus a few additional

ideas, we obtain a polynomial-time $3/2+eps$-approximation for 2DKR, which improves to $4/3+eps$ in the cardinality case.

This is a joint work with Waldo Galvez, Fabrizio Grandoni, Sandy Heydrich,

Salvatore Ingala and Andreas Wiese. This result will appear in FOCS 2017.

Biography of the speaker

Arindam Khan is a researcher in IDSIA, SUPSI in Lugano, Switzerland. His

research areas include approximation algorithms, online algorithms and

computational geometry. He has obtained his PhD in Algorithms, Combinatorics

and Optimization (ACO) from Georgia Institute of Technology, Atlanta, USA under

Prof. Prasad Tetali. Previously he has been a research intern in Theory group,

Microsoft Research Redmond and Microsoft Research Silicon Valley USA, a visiting

researcher at Simons Institute, Berkeley, USA and a blue scholar in IBM Research

India.

Host Faculty : Dr. Anand Louis

ALL ARE WELCOME]]>

Ph.D. Thesis Defense

Speaker : Mr. Cherukapally Srikanth

Ph.D student

Dept. of CSA

Title : Number Theoretic, Computational and

Cryptographic Aspects of a Certain Sequence of Arithmetic Progressions

Faculty Advisor : Prof. C.E. Veni Madhavan & Dr. Sanjit Chatterjee

Date : Wednesday, August 09, 2017

Time : 11:00 AM

Venue : CSA Seminar Hall (Room No. 254, First Floor)

Live Video : MMCRs - ECE, CEDT, EE

or

URL -- http://mmcr.iisc.ernet.in:8008/live/csa.html

Abstract

This thesis introduces a new mathematical object: collection of

arithmetic progressions with elements satisfying the inverse property,

``j-th terms of i-th and (i+1)-th progressions are multiplicative

inverses of each other modulo (j+1)-th term of i-th progression''.

Such a collection is uniquely defined for any pair $(a,d)$ of co-prime integers.

The progressions of the collection are ordered. Thus we call it a sequence rather

than a collection. The results of the thesis are on the following number theoretic, computational

and cryptographic aspects of the defined sequence and its generalizations.

The sequence is closely connected to the classical Euclidean

algorithm. Precisely, certain consecutive progressions of the sequence

form ``groupings''. The difference between the common differences of

any two consecutive progressions of a grouping is same. The number of

progressions in a grouping is connected to the quotient sequence

of the Euclidean algorithm on co-prime input pairs. The research community

has studied extensively the behavior of the Euclidean algorithm. For the first

time in the literature, the connection (proven in the thesis) shows

what the quotients of the algorithm signify. Further, the leading terms of progressions

within groupings satisfy a mirror image symmetry property,

called ``symmetricity''. The property is subject to the quotient

sequence of the Euclidean algorithm and

divisors of integers of the form $x^2-y^2$ falling in specific intervals.

The integers $a$, $d$ are the primary quantities of the defined sequence in a computational sense.

Given the two, leading term and common difference of any progression of the sequence

can be computed in time quadratic in the binary length of $d$.

On the other hand, the inverse computational question of

finding $(a,d)$, given information on some terms of the sequence, is interesting.

This problem turns out to be hard as it requires finding solutions to

an nearly-determined system of multivariate polynomial equations. Two sub-problems

arising in this context are shown to be equivalent to the problem of factoring integers. The reduction

to the factoring problem, in both cases, is probabilistic.

Utilizing the computational difficulty of solving the inverse problem, and the sub-problems (mentioned above),

we propose a symmetric-key cryptographic scheme (SKCS),

and a public key cryptographic scheme (PKCS). The PKCS is also based on the hardness of the problem of

finding square-roots modulo composite integers. Our proposal uses the same

algorithmic and computational primitives

for effecting both the PKCS and SKCS. In addition, we use the notion of

the sequence of arithmetic progressions to design an entity authentication scheme.

The proof of equivalence between one of the inverse computational problems (mentioned above)

and integer factoring led us to formulate and investigate

an independent problem concerning the largest divisor of integer $N$ bounded by

the square-root of $N$. We present some algorithmic and combinatorial results.

In the course of the above investigations, we are led to certain open questions of number theoretic,

combinatorial and algorithmic nature. These pertain to the quotient sequence of the Euclidean algorithm,

divisors of integers of the form $x^2-y^2$ in specific intervals,

and the largest divisor of integer $N$ bounded by its square-root.

ALL ARE WELCOME]]>

Department of Computer Science and Automation

CSA Distinguished Lecture

Speaker : Professor Krishna R. Pattipati

University of Connecticut

Storrs, Connecticut

USA

Title : MULTI-OBJECTIVE NAVIGATION IN UNCERTAINTY

Date : Thursday, August 03, 2017

Time : 3:00 PM

Venue : CSA Seminar Hall (Room No. 254, First Floor)

Live Video : MMCRs - ECE, CEDT, EE

or

URL -- http://mmcr.iisc.ernet.in:8008/live/csa.html

Abstract

Routing in uncertain environments involves many contextual elements, such as different environmental

conditions (ensemble forecasts with varying spatial and temporal uncertainty), multiple objectives,

changes in mission goals while en route (e.g., training requirements, pop-up threats, humanitarian aid),

and asset status. In this walk, we focus on robust planning under uncertainty by exploiting weather

forecast ensembles and realizations using TMPLAR, a Tool for Multi-objective PLanning and Asset Routing,

in the context of 3-D and 4-D path navigation. Our approach includes solving for m-best shortest paths

for each weather forecast realization via Murty's search space partitioning strategy and evaluating

the mean, variance, and signal-to-noise ratio (SNR) of the paths over all weather possibilities.

The robust path is one that, for example, minimizes the root mean square (RMS) value or one that

maximizes the SNR, given the possible forecast realizations. In a complementary effort to compact the

feature-rich multi-dimensional navigational problem space, we develop a self-organizing map

(SOM)-based data reduction scheme that filters complex contexts and highlights only the key

impacting features within a given information space. We demonstrate the utility of our approaches

via application to a real-world shipping tragedy, using the weather forecast realizations available prior to the event.

Biography of the speaker

Krishna R. Pattipati received the B. Tech. degree in electrical engineering with

highest honours from the Indian Institute of Technology, Kharagpur, in 1975, and

the M.S. and Ph.D. degrees in control and communication systems from UConn,

Storrs, in 1977 and 1980, respectively. He was with ALPHATECH, Inc.,

Burlington, MA from 1980 to 1986. He has been with the Department of Electrical

and Computer Engineering at UConn since 1986, where he is currently the Board of

Trustees Distinguished Professor and the UTC Chair Professor in Systems

Engineering. Dr. Pattipati’s research activities are in the areas of

proactive decision support, uncertainty quantification, smart manufacturing,

autonomy, knowledge representation, and optimization-based learning and

inference. A common theme among these applications is that they are

characterized by a great deal of uncertainty, complexity, and computational

intractability. He is a cofounder of Qualtech Systems, Inc., a firm

specializing in advanced integrated diagnostics software tools (TEAMS, TEAMS-RT,

TEAMS-RDS, TEAMATE, PackNGo), and serves on the board of Aptima, Inc.

Dr. Pattipati was selected by the IEEE Systems, Man, and Cybernetics (SMC)

Society as the Outstanding Young Engineer of 1984, and received the Centennial

Key to the Future award. He has served as the Editor-in-Chief of the IEEE

TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS--PART B from 1998 to 2001. He was

co-recipient of the Andrew P. Sage Award for the Best SMC Transactions Paper for

1999, the Barry Carlton Award for the Best AES Transactions Paper for 2000, the

2002 and 2008 NASA Space Act Awards for "A Comprehensive Toolset for Model-based

Health Monitoring and Diagnosis," and “Real-time Update of Fault-Test

Dependencies of Dynamic Systems: A Comprehensive Toolset for Model-Based Health

Monitoring and Diagnostics”, the 2003 AAUP Research Excellence Award and the

2005 School of Engineering Teaching Excellence Award at the University of

Connecticut at UCONN. He is an elected Fellow of IEEE for his contributions to

discrete-optimization algorithms for large-scale systems and team

decision-making, and is a member of the Connecticut Academy of Science and

Engineering.

Host Faculty : Prof. Y Narahari

ALL ARE WELCOME]]>

CSA Distinguished Lecture

Speaker : Professor Krishna R. Pattipati

University of Connecticut

Storrs, Connecticut

USA

Title : MULTI-OBJECTIVE NAVIGATION IN UNCERTAINTY

Date : Wednesday, August 02, 2017

Time : 4:00 PM

Venue : CSA Seminar Hall (Room No. 254, First Floor)

Live Video : MMCRs - ECE, CEDT, EE

or

URL -- http://mmcr.iisc.ernet.in:8008/live/csa.html

Abstract

Routing in uncertain environments involves many contextual elements, such as different environmental

conditions (ensemble forecasts with varying spatial and temporal uncertainty), multiple objectives,

changes in mission goals while en route (e.g., training requirements, pop-up threats, humanitarian aid),

and asset status. In this walk, we focus on robust planning under uncertainty by exploiting weather

forecast ensembles and realizations using TMPLAR, a Tool for Multi-objective PLanning and Asset Routing,

in the context of 3-D and 4-D path navigation. Our approach includes solving for m-best shortest paths

for each weather forecast realization via Murty's search space partitioning strategy and evaluating

the mean, variance, and signal-to-noise ratio (SNR) of the paths over all weather possibilities.

The robust path is one that, for example, minimizes the root mean square (RMS) value or one that

maximizes the SNR, given the possible forecast realizations. In a complementary effort to compact the

feature-rich multi-dimensional navigational problem space, we develop a self-organizing map

(SOM)-based data reduction scheme that filters complex contexts and highlights only the key

impacting features within a given information space. We demonstrate the utility of our approaches

via application to a real-world shipping tragedy, using the weather forecast realizations available prior to the event.

Biography of the speaker

Krishna R. Pattipati received the B. Tech. degree in electrical engineering with

highest honours from the Indian Institute of Technology, Kharagpur, in 1975, and

the M.S. and Ph.D. degrees in control and communication systems from UConn,

Storrs, in 1977 and 1980, respectively. He was with ALPHATECH, Inc.,

Burlington, MA from 1980 to 1986. He has been with the Department of Electrical

and Computer Engineering at UConn since 1986, where he is currently the Board of

Trustees Distinguished Professor and the UTC Chair Professor in Systems

Engineering. Dr. Pattipati’s research activities are in the areas of

proactive decision support, uncertainty quantification, smart manufacturing,

autonomy, knowledge representation, and optimization-based learning and

inference. A common theme among these applications is that they are

characterized by a great deal of uncertainty, complexity, and computational

intractability. He is a cofounder of Qualtech Systems, Inc., a firm

specializing in advanced integrated diagnostics software tools (TEAMS, TEAMS-RT,

TEAMS-RDS, TEAMATE, PackNGo), and serves on the board of Aptima, Inc.

Dr. Pattipati was selected by the IEEE Systems, Man, and Cybernetics (SMC)

Society as the Outstanding Young Engineer of 1984, and received the Centennial

Key to the Future award. He has served as the Editor-in-Chief of the IEEE

TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS--PART B from 1998 to 2001. He was

co-recipient of the Andrew P. Sage Award for the Best SMC Transactions Paper for

1999, the Barry Carlton Award for the Best AES Transactions Paper for 2000, the

2002 and 2008 NASA Space Act Awards for "A Comprehensive Toolset for Model-based

Health Monitoring and Diagnosis," and “Real-time Update of Fault-Test

Dependencies of Dynamic Systems: A Comprehensive Toolset for Model-Based Health

Monitoring and Diagnostics”, the 2003 AAUP Research Excellence Award and the

2005 School of Engineering Teaching Excellence Award at the University of

Connecticut at UCONN. He is an elected Fellow of IEEE for his contributions to

discrete-optimization algorithms for large-scale systems and team

decision-making, and is a member of the Connecticut Academy of Science and

Engineering.

Host Faculty : Prof. Y Narahari

ALL ARE WELCOME]]>

Department Seminar

Speaker : Prof. Subramanian Ramamoorthy

Title : The program induction route to explainable AI

Date : Wednesday, August 02, 2017

Time : 11:30 AM

Venue : CSA Seminar Hall (Room No. 254, First Floor)

Live Video : MMCRs - ECE, CEDT, EE

or

URL -- http://mmcr.iisc.ernet.in:8008/live/csa.html

Abstract

The confluence of advances in diverse areas including machine learning, large scale

computing and reliable commoditized hardware have brought autonomous robots to the

point where they are poised to be genuinely a part of our daily lives. Some of the

application areas where this seems most imminent, e.g., autonomous vehicles, also

bring with them stringent requirements regarding safety, explainability and

trustworthiness. These needs seem to be at odds with the ways in which recent

successes have been achieved, e.g., with end-to-end learning. In this talk, I will

try to make a case for an approach to bridging this gap, through the use of programmatic

representations that intermediate between opaque but efficient learning methods and

other techniques for reasoning that benefit from ‘symbolic’ representations.

I will start by attempting to frame the overall problem, drawing on some of the

motivations of the DARPA Explainable AI program (under the auspices of which we are

starting a new project, COGLE) and on extant ideas regarding safety and dynamical

properties in the control theorists’ toolbox.

Our first result will be based on a framework for Grounding and Learning Instances

through Demonstration and Eye tracking (GLIDE). The problem here is to learn the

mapping between abstract plan symbols and their physical instances in the environment,

i.e., physical symbol grounding, starting from cross-modal input provides the combination

of high- level task descriptions (e.g., from a natural language instruction) and a

detailed video or joint angles signal. This problem is formulated in terms of a

probabilistic generative model and addressed using an algorithm for computationally

feasible inference to associate traces of task demonstration to a sequence of fixations

which we call fixation programs.

The second related result more directly addresses the issue of inferring processes that

underlie observed black-box phenomena, in the form of causal mechanisms. We propose to

learn high-level programs in order to represent abstract models, which capture the

invariant structure in the observed data. We introduce the pi-machine (program-induction

machine) – an architecture able to induce interpretable LISP-like programs from observed

data traces. We propose an optimization procedure for program learning based on back

propagation, gradient descent and A* search. We apply the proposed method to problems

including system identification, explaining the behavior of a DQN agent and explaining

human demonstrations in planned activities. Our results show that the pi-machine can

efficiently induce interpretable programs from individual data traces.

Biography of the speaker

Prof. Subramanian Ramamoorthy is a Reader (Associate Professor) in the School of

Informatics, University of Edinburgh, where he has been on the faculty since

2007. He is a Coordinator of the EPSRC Robotarium Research Facility, Executive

Committee Member for the Edinburgh Centre for Robotics and at the Bayes Centre.

He received his PhD in Electrical and Computer Engineering from The University

of Texas at Austin in 2007. He is an elected Member of the Young Academy of

Scotland at the Royal Society of Edinburgh, and has been a Visiting Professor at

Stanford University and the University of Rome “La Sapienza”. His research

focus is on robot learning and decision-making under uncertainty, addressed

through a combination machine learning techniques with emphasis on issues of

transfer, online and reinforcement learning as well as new representations and

analysis techniques based on geometric/topological abstractions.

Host Faculty : Aditya Kanade

ALL ARE WELCOME]]>