Murray Wiki - User contributions [en]

Michaëlle Mayalu, Jan 2017

2017-01-21T00:20:39Z

Nmatni: /* 23 Jan (Mon) */

Michaëlle Mayalu from MIT will visit on 23-24 January. __NOTOC__

=== Schedule ===
{| border=1
|- valign=top
| width=50% |
==== 23 Jan (Mon) ====
* 8 am: Breakfast with Richard (Ath)
* 9 am: Anandh (103 Steele?)
* 10 am: DARPA Biological Control program meeting, 168 Broad (walk over with Richard)
<hr>
* 12 pm: Seminar, 121 Annenberg (abtract below)
* 1 pm: lunch with TBD
<hr>
* 2 pm: Open
* 2:45 pm: Noah Olsman (2nd floor lounge, Annenberg)
* 3:30 pm: Nikolai Matni (2nd floor lounge, Annenberg)
* 4:15 pm: Aaron Ames, 266 Gates-Thomas
<hr>
* 5:30 pm: Meet Richard in 107 Steele. Walk to dinner.

| width=50% |

==== 24 Jan (Tue) ====
* 9:00 am: Open
* 9:45 am: Open
* 10:30 am: Cindy (2nd floor lounge, Annenberg)
* 11:15 am: Ania (2nd floor lounge, Annenberg)
<hr>
* 12 pm: Henrike Niederholtmeyer seminar, 121 Annenberg
<hr>
* 1:00 pm: Open
* 1:45 pm: [[Jan 2017 meeting schedule|Richard]]
* 2:30 pm: John Doyle
* 3:15 pm: Vipul (Keck 222)
* 4:00 pm: Open
* 4:45 pm: Open
* 5:30 pm: Done for the day
|}

=== Seminar info ===
<center>
'''Reduced-order Systems Approach to Prediction of Emergent Behaviors'''

Michaëlle N. Mayalu 
Massachusetts Institute of Technology

Monday, January 23, 2017 
12-1 pm, 121 Annenberg
</center>

Increased understanding of the physical and chemical principles that drive a biological process has led to the development of mathematical formulations that are used to simulate a rich variety of responses. As a result, a vast amount of simulation data can be created for analyzing single cell behavior numerically. However complex and extensive mechanisms involved in emergent behavior of multiple interacting cells may become intractable due to mathematical and computational complexity. This talk will address how we can exploit simulation data describing the nonlinear dynamics of single cell behavior to create a reduced- order linear state equation in latent variable space. Furthermore, the linearity of the reduced-order latent variable state equation allows for the superposition of multiple solutions to predict emergent behaviors of interacting cells.
The linear latent state equation describing the nonlinear dynamics of single cell is created in two steps. First the original independent state variables are augmented by adding auxiliary variables necessary to “sufficiently inform” the single cell nonlinear dynamics. This creates a high-dimensional state space where a linear description of the nonlinear system can be found. Second, latent variables extracted from the high-dimensional state space and used to create the reduced-order linear equation. While the resultant latent state equation is linear, complex nonlinearities are embedded in the compact model, leading to precise and global linearization of nonlinear dynamics. Furthermore, in order to predict multi-cell emergent behavior, the reduced-order linear models of single cells are used as agents in a comprehensive agent-based framework based on linear superposition of mutually shared variables. The approach is motivated by emergent behaviors in collective cell migration in order to gain insight for the study and control cancer metastasis and wound healing. However, the general approach may be applied to systems of interacting nonlinear agents, which would otherwise be prohibitively complex to compute.

Michaëlle N. Mayalu is a Ph.D. student at the Brit and Alex d’Arbeloff Laboratory for Information Systems and Technology in the Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge MA. She received the B.S., and M.S., degrees in Mechanical Engineering in 2010 and 2012 from Massachusetts Institute of Technology. Her thesis work is focused on modeling and predicting biological systems behavior by drawing on aspects of dynamic modeling and simulation, data analysis, statistical learning and control theory. Her thesis supervisor is Professor H. Harry Asada in the Department of Mechanical Engineering at MIT.

Lillian Ratliff, 1-2 Feb 2016

2016-02-01T18:59:53Z

Nmatni: /* Schedule */

Lillian Ratfliff will be visiting Caltech on 1-2 Feb 2016. Please sign up below if you would like to meet with her.

=== Schedule ===

1 Feb 2016 (Mon)
* Noon: arrive at BUR. Travel to Pasadena (Caltech car)
* 1 pm: Yisong (lunch) - meet at the Ath
* 2 pm: Pietro Perona, 104A Moore
* 2:45 pm: open
* 3:30 pm: open
* 4:00 pm: Seminar - Munzer Dahleh, 105 Annenberg
* 5 pm: done for the day
* 6:30 pm: meet in the lobby of the Ath for dinner with Nik Matni, Vanessa Jonsson, Dj Krishanmurth (location TBD)

2 Feb 2016 (Tue)
* 7:45 am: breakfast with Richard (Ath)
* 8:45 am: open
* 9:30 am: Dj, Annenberg 218
* 10:15 am: Adam W, 215 Annenberg
* 11:00 am: Nikolai M, Annenberg 2nd floor lounge
* 11:45 am: seminar setup
* 12:00 pm: seminar, 104 ANB
* 1 pm: lunch with Richard (Chandler)
* 1:45 pm: Vanessa Jonsson (330 Braun)
* 2:30 pm: Joel Tropp, 307 Annenberg
* 3:15 pm: Niangjun Chen
* 4:00 pm: Ivan Papusha, 231 Annenberg
* 4:45 pm: open
* 5:30 pm: done for the day
* 6:15 pm: depart from Ath for BUR (Caltech car)

=== Seminar: 2 Feb (Wed), 12-1 pm, 105 ANB ===

TITLE:
The Emerging Data Market: Adaptive Incentives for Smart, Connected Infrastructure

ABSTRACT:
The next generation urban ecosystem empowered by the internet of things has at its core a shared economy where physical resources and data are easily aggregated and exchanged. In particular, advances in technology have lead to the proliferation of smart devices that provide access to streaming data and platforms for novel sharing mechanisms. This has, in turn, resulted in an emerging marketplace in which data is a commodity. At the same time, many urban constituents are increasingly becoming aware of the value of their data and its usefulness for operations. In such an environment, new learning and optimization schemes which consider users as strategic data sources and resource seekers are needed. In this talk, we will discuss the emerging data market, its incentive structure (players and their motivations), and tools for learning with strategic data sources. Focusing on the design of adaptive incentive mechanisms under adverse selection, we will construct an algorithm for online utility learning and incentive design and show convergence results for both the case where players are rational (play according to Nash) and myopic. We will see through a tutorial example how the algorithm performs, and conclude with some open questions and future directions.

Lillian Ratliff, 1-2 Feb 2016

2016-01-28T21:40:48Z

Nmatni: /* Schedule */

Lillian Ratfliff will be visiting Caltech on 1-2 Feb 2016. Please sign up below if you would like to meet with her.

=== Schedule ===

1 Feb 2016 (Mon)
* Noon: arrive at BUR. Travel to Pasadena (Caltech car)
* 1 pm: Yisong (and John Doyle?) (lunch)
* 2 pm: Pietro Perona
* 2:45 pm: open
* 3:30 pm: open
* 4:15 pm: open
* 5 pm: done for the day
* 6 pm: optional dinner with graduate students and postdocs

2 Feb 2016 (Tue)
* 7:45 am: breakfast with Richard (Ath)
* 8:45 am: open
* 9:30 am: open
* 10:15 am: Adam W
* 11:00 am: Nikolai M
* 11:45 am: seminar setup
* 12:00 pm: seminar, 104 ANB
* 1 pm: lunch with Richard (Chandler)
* 1:45 pm: open
* 2:30 pm: Joel Tropp
* 3:15 pm: open
* 4:00 pm: open
* 4:45 pm: open
* 5:30 pm: done for the day
* 6:15 pm: depart from Ath for BUR (Caltech car)

=== Seminar: 2 Feb (Wed), 12-1 pm, 105 ANB ===

TITLE:
The Emerging Data Market: Adaptive Incentives for Smart, Connected Infrastructure

ABSTRACT:
The next generation urban ecosystem empowered by the internet of things has at its core a shared economy where physical resources and data are easily aggregated and exchanged. In particular, advances in technology have lead to the proliferation of smart devices that provide access to streaming data and platforms for novel sharing mechanisms. This has, in turn, resulted in an emerging marketplace in which data is a commodity. At the same time, many urban constituents are increasingly becoming aware of the value of their data and its usefulness for operations. In such an environment, new learning and optimization schemes which consider users as strategic data sources and resource seekers are needed. In this talk, we will discuss the emerging data market, its incentive structure (players and their motivations), and tools for learning with strategic data sources. Focusing on the design of adaptive incentive mechanisms under adverse selection, we will construct an algorithm for online utility learning and incentive design and show convergence results for both the case where players are rational (play according to Nash) and myopic. We will see through a tutorial example how the algorithm performs, and conclude with some open questions and future directions.

Gupta Visit

2014-05-06T01:21:36Z

Nmatni:

Abhishek Gupta is a PhD candidate in Aerospace Engineering department at UIUC, where he works with Cedric Langbort and Tamer Basar on theoretical problems in Stochastic Control theory, Game Theory, Optimization and Information theory.

He is going to be giving a talk on Thursday May 8th, at 10am in Annenberg 213 on the existence of optimal strategies for dynamic games and teams with asymmetric information.

He will be around for most of Thursday, so please sign up below (name and location) if you would like to meet with him.

Schedule:
{| border=1
|-
|''' Time''' || '''Person'''
|-
| 8:30a || Arrive at Caltech
|-
| 9:00a || Nikolai - Annenberg Lounge
|-
| 9:30a || Open
|-
| 10:00a || Seminar - 213 Annenberg
|-
| 10:30a || Seminar - 213 Annenberg
|-
| 11:00a || Richard, 109 Steele Lab (meeting in Annenberg after seminar)
|-
| 11:30a || Open
|-
| 12:00p || Lunch
|-
| 1:30p || Open
|-
| 2:00 p || Doyle 210 Annenberg
|-
| 2:30 p || Depart for LAX
|}

Gupta Visit

2014-05-05T21:21:40Z

Nmatni:

Abhishek Gupta is a PhD candidate in Aerospace Engineering department at UIUC, where he works with Cedric Langbort and Tamer Basar on theoretical problems in Stochastic Control theory, Game Theory, Optimization and Information theory.

He is going to be giving a talk on Thursday May 8th, at 10am in Annenberg 213 on the existence of optimal strategies for dynamic games and teams with asymmetric information.

He will be around for most of Thursday, so please sign up below (name and location) if you would like to meet with him.

Schedule:
{| border=1
|-
|''' Time''' || '''Person'''
|-
| 8:30a || Arrive at Caltech
|-
| 9:00a || Open
|-
| 9:30a || Open
|-
| 10:00a || Seminar - 213 Annenberg
|-
| 10:30a || Seminar - 213 Annenberg
|-
| 11:00a || Open
|-
| 12:00p || Lunch
|-
| 1:30p || Open
|-
| 2:00 p || Open
|-
| 2:30 p || Depart for LAX
|}

Gupta Visit

2014-05-05T21:20:37Z

Nmatni:

Abhishek Gupta is a PhD candidate in Aerospace Engineering department at UIUC, where he works with Cedric Langbort and Tamer Basar on theoretical problems in Stochastic Control theory, Game Theory, Optimization and Information theory.

He is going to be giving a talk on Thursday May 8th, at 10am in Annenberg 213 on the existence of optimal strategies for dynamic games and teams with asymmetric information.

He will be around for most of Thursday, so please sign up below (name and location) if you would like to meet with him.

Schedule:
{| border=1
|-
|''' Time''' || '''Person'''
|-
| 10:00a || Seminar - 213 Annenberg
|-
| 10:30a || Seminar - 213 Annenberg
|-
| 11:00a || Open
|-
| 12:00p || Lunch
|-
| 1:30p || Open
|-
| 2:00 p || Open
|-
| 2:30 p || Depart for LAX
|}

Gupta Visit

2014-05-05T20:03:39Z

Nmatni:

Abhishek Gupta is a PhD candidate in Aerospace Engineering department at UIUC, where he works with Cedric Langbort and Tamer Basar on theoretical problems in Stochastic Control theory, Game Theory, Optimization and Information theory.

He is going to be giving a talk on Thursday May 8th, at 10am in Annenberg 213 on the existence of optimal strategies for dynamic games and teams with asymmetric information.

He will be around for most of Thursday, so please sign up below (name and location) if you would like to meet with him.

Schedule:
{| border=1
|-
|''' Time''' || '''Person'''
|-
| 10:00a || Seminar - 213 Annenberg
|-
| 10:30a || Seminar - 213 Annenberg
|-
| 11:00a || Open
|-
| 12:00p || Lunch
|-
| 1:30p || Open
|-
| 2:00 p || Open
|-
| 2:30 p || Open
|-
| 3:00 p || Open
|-
| 3:30 p || Open
|}

Gupta Visit

2014-05-05T20:01:56Z

Nmatni:

Gupta Visit

2014-05-05T20:01:30Z

Nmatni:

Gupta Visit

2014-05-05T20:01:13Z

Nmatni:

Alex Mauroy, Nov 2013

2013-11-21T16:39:55Z

Nmatni: /* Schedule */

=== Schedule ===

* 9:30 am: meet with Richard, 109 Steele
* 10:00 am: Open
* 10:30 am: Open
* 11:00 am: Prof. Oscar Bruno
* 11:30 am: Seminar prep
* 12:00 pm: Group Meeting Seminar, 213 Annenberg
* 1:15 pm: Meet with John Doyle
* 2:00 pm: Seungil You, 231 Annenberg
* 2:45 pm: Eric Wolff, 331 Annenberg
* 3:30 pm: Nikolai Matni, 335 Annenberg
* 4:15 pm: Open

=== Talk Abstract===

===== A spectral operator-theoretic approach to dissipative systems: geometry and global stability =====

Operator-theoretic methods provide a powerful global description of dynamical systems. They transform a nonlinear system into a linear (but infinite-dimensional) system and are therefore well-suited to spectral analysis. These methods have been largely developed over the past decades for conservative systems (i.e., on an attractor), in the context of unitary operators, but they have almost never been applied to dissipative systems (i.e., off the attractor).
> In this talk, we present an operator-theoretic framework for the study of dissipative systems, which relies on the so-called Koopman operator. In particular, we show that the global geometric properties of the systems are captured by the spectral properties of the Koopman operator. For instance the well-known isochrons of limit cycles are related to an eigenfunction of the operator, an observation which yields a new method for computing them and which is used to extend the concept to other types of attractors. In addition, (global) stability properties of the system are revealed by the existence of particular eigenfunctions. This provides a new framework for the global stability analysis of general attractors.

SoCal Control Workshop Carpool 2012-11-30

2012-11-28T19:59:12Z

Nmatni: /* Returning to Pasadena */

Please enter information in the table below if you can give a ride or need a ride. If you can't edit the table, send e-mail to the driver (if you need a ride) or to Richard (if you can provide a ride).

* [http://tintoretto.ucsd.edu/jorge/sccw/index.html Workshop web page]

{| width=100% cellpadding=0 cellspacing=0
|- valign=top
| width=50% align=left|
==== Driving out to UCSD ====
{| border=1
|- valign=top
| Departure time
| Driver
| Passengers
|- valign=top
| 6:30 am
| Richard Murray murray@cds
| 3 passengers:
* Vipul
* Yoke Peng
* Eric
|- valign=top
| 6:30 am
| Marcella mgomez@caltech
| 3 passengers:
* Ioannis
* Zoltan
* Mickey
|- valign=top
| 6:30 am
| Kevin Shan kqshan@gmail
| 3 passengers:
* Seungil
* Anandh
*
|- valign=top
| 6:20 am
| Yanan Sui ysui@caltech
| 3 passengers:
*
*
*
|}
| width=50% align=left|

==== Returning to Pasadena ====
{| border=1
|- valign=top
| Departure time
| Driver
| Passengers
|- valign=top
| 5:30 pm
| Richard Murray
| 3 passengers
* Vipul
* Yoke Peng
* Eric
|- valign=top
| After dinner ~7:30 pm
| Marcella
| 3 passengers
* Ioannis
* Zoltan
* Mickey
|- valign=top
| 5:30pm
| Kevin Shan
| 3 passengers
* Seungil
* Anandh
*
|- valign=top
| 5:30pm
| Yanan Sui
| 3 passengers
*
*
*
|}

SoCal Control Workshop Carpool 2012-11-30

2012-11-28T19:58:58Z

Nmatni: /* Driving out to UCSD */

SoCal Control Workshop Carpool 2012-11-30

2012-11-28T00:50:40Z

Nmatni: /* Returning to Pasadena */

Please enter information in the table below if you can give a ride or need a ride. If you can't edit the table, send e-mail to the driver (if you need a ride) or to Richard (if you can provide a ride).

* [http://tintoretto.ucsd.edu/jorge/sccw/index.html Workshop web page]

{| width=100% cellpadding=0 cellspacing=0
|- valign=top
| width=50% align=left|
==== Driving out to UCSD ====
{| border=1
|- valign=top
| Departure time
| Driver
| Passengers
|- valign=top
| 6:30 am
| Richard Murray
| 3 passengers:
* Vipul
* Yoke Peng
* Eric
|- valign=top
| 6:30 am
| Marcella (mgomez@caltech.edu)
| 3 passengers:
* Ioannis
* Zoltan
* Mickey
|- valign=top
| 6:30 am
| Kevin Shan
| 3 passengers:
* Seungil
* Anandh
* Stephanie
|- valign=top
| 6:30 am
|
| 3 passengers:
* Nikolai
*
*
|}
| width=50% align=left|

==== Returning to Pasadena ====
{| border=1
|- valign=top
| Departure time
| Driver
| Passengers
|- valign=top
| 5:30 pm
| Richard Murray
| 3 passengers
* Vipul
* Yoke Peng
* Eric
|- valign=top
| After dinner ~7:30 pm
| Marcella
| 3 passengers
* Ioannis
* Zoltan
|- valign=top
| 5:30pm
| Kevin Shan
| 3 passengers
* Seungil
* Anandh
* Stephanie
|- valign=top
| 5:30pm
|
| 3 passengers
* Nikolai
*
*
|}

SoCal Control Workshop Carpool 2012-11-30

2012-11-28T00:50:04Z

Nmatni: /* Driving out to UCSD */

Please enter information in the table below if you can give a ride or need a ride. If you can't edit the table, send e-mail to the driver (if you need a ride) or to Richard (if you can provide a ride).

* [http://tintoretto.ucsd.edu/jorge/sccw/index.html Workshop web page]

{| width=100% cellpadding=0 cellspacing=0
|- valign=top
| width=50% align=left|
==== Driving out to UCSD ====
{| border=1
|- valign=top
| Departure time
| Driver
| Passengers
|- valign=top
| 6:30 am
| Richard Murray
| 3 passengers:
* Vipul
* Yoke Peng
* Eric
|- valign=top
| 6:30 am
| Marcella (mgomez@caltech.edu)
| 3 passengers:
* Ioannis
* Zoltan
* Mickey
|- valign=top
| 6:30 am
| Kevin Shan
| 3 passengers:
* Seungil
* Anandh
* Stephanie
|- valign=top
| 6:30 am
|
| 3 passengers:
* Nikolai
*
*
|}
| width=50% align=left|

==== Returning to Pasadena ====
{| border=1
|- valign=top
| Departure time
| Driver
| Passengers
|- valign=top
| 5:30 pm
| Richard Murray
| 3 passengers
* Vipul
* Yoke Peng
* Eric
|- valign=top
| After dinner ~7:30 pm
| Marcella
| 3 passengers
* Ioannis
* Zoltan
|- valign=top
| 5:30pm
| Kevin Shan
| 3 passengers
* Seungil
* Anandh
* Stephanie
|}

Henrik Sandberg, July 2012

2012-07-28T00:09:32Z

Nmatni: /* Monday */

Henrik Sandberg will be visiting CDS on 30-31 July (Tue-Wed). __NOTOC__

=== Schedule ===
{| width=100% border=1
|- valign=top
| width=50% |
==== Monday ====
* 10 am: Open
* 11 am: Seminar, 213 Annenberg
* 12 noon: Lunch, hosted by Nikolai
<hr>
* 1:30 pm: Matanya Horowitz & Nikolai
* 2:15 pm: Shuo Han
* 3:00 pm: Open
* 3:45 pm: KISS workshop on ultra-reliability
|

==== Tuesday ====
* 10 am: Enoch
* 10:45 am: Eric Wolff
* 11:30 am: Open
<hr>
* 1:30 pm: Aristo
* 2:15 pm: Open
* 3:00 pm: Open
* 3:45 pm: Open
* 4:30 pm: Open
<hr>
* 6:30 pm: Richard
|}

=== Abstract ===

'''Synthesis of Models using Lossless Components and Some Related Design Trade-Offs'''

Speaker: Henrik Sandberg 
Affiliation: Electrical Engineering, KTH 
Date and time: Monday, 30 July 2012 - 11:00am 
Location: 213 Annenberg 

The area of network synthesis deals with how to implement mathematical models using a well-defined set of physical components. This area was thoroughly studied in the circuits community in the 1960-1970's, using components such as resistors, transformers, capacitors, inductors, and operational amplifiers. Today similar questions are being asked in synthetic biology and in mechanics, only now the given sets of physical components are very different. Typical questions of interest are to characterize what type of mathematical models can be implemented for given sets of components, and the minimal number of components required for a specific model.

In this talk, we address the problem of synthesizing some classes of active and passive mathematical models using only a small number of lossless physical components. The set of lossless components are relevant since they conserve energy and if we intend to implement anything on a microscopic level, for example a micro machine, we will need to respect this limitation. In our earlier work we have shown that linear models are passive if, and only if, they can be approximated to any desired degree of accuracy by a linear lossless system, over any fixed time horizon. Typically this requires a huge number of lossless components, however, which results in thermal noise as quantified by the fluctuation-dissipation theorem. Here we will see that by allowing the linear lossless components to be time-varying, not only can we avoid some thermal noise, but we can also synthesize some active components. Given the importance of active components such as operational amplifiers this is encouraging. The price for the decrease in implementation complexity is a need for an external power supply. We illustrate the control-theoretic relevance of these results by means of a few examples and design trade-offs.

Henrik Sandberg, July 2012

2012-07-27T18:44:25Z

Nmatni: /* Monday */

Henrik Sandberg will be visiting CDS on 30-31 July (Tue-Wed). __NOTOC__

=== Schedule ===
{| width=100% border=1
|- valign=top
| width=50% |
==== Monday ====
* 10 am: Open
* 11 am: Seminar, 213 Annenberg
* 12 noon: Lunch, hosted by Nikolai
<hr>
* 1:30 pm: Matanya Horowitz
* 2:15 pm: Shuo Han
* 3:00 pm: Nikolai Matni
* 3:45 pm: Open
* 4:30 pm: Open
|

==== Tuesday ====
* 10 am: Enoch
* 10:45 am: Open
* 11:30 am: Open
<hr>
* 1:30 pm: Aristo
* 2:15 pm: Open
* 3:00 pm: Open
* 3:45 pm: Open
* 4:30 pm: Open
<hr>
* 6:30 pm: Richard
|}

=== Abstract ===

'''Synthesis of Models using Lossless Components and Some Related Design Trade-Offs'''

Speaker: Henrik Sandberg 
Affiliation: Electrical Engineering, KTH 
Date and time: Monday, 30 July 2012 - 11:00am 
Location: 213 Annenberg 

The area of network synthesis deals with how to implement mathematical models using a well-defined set of physical components. This area was thoroughly studied in the circuits community in the 1960-1970's, using components such as resistors, transformers, capacitors, inductors, and operational amplifiers. Today similar questions are being asked in synthetic biology and in mechanics, only now the given sets of physical components are very different. Typical questions of interest are to characterize what type of mathematical models can be implemented for given sets of components, and the minimal number of components required for a specific model.

In this talk, we address the problem of synthesizing some classes of active and passive mathematical models using only a small number of lossless physical components. The set of lossless components are relevant since they conserve energy and if we intend to implement anything on a microscopic level, for example a micro machine, we will need to respect this limitation. In our earlier work we have shown that linear models are passive if, and only if, they can be approximated to any desired degree of accuracy by a linear lossless system, over any fixed time horizon. Typically this requires a huge number of lossless components, however, which results in thermal noise as quantified by the fluctuation-dissipation theorem. Here we will see that by allowing the linear lossless components to be time-varying, not only can we avoid some thermal noise, but we can also synthesize some active components. Given the importance of active components such as operational amplifiers this is encouraging. The price for the decrease in implementation complexity is a need for an external power supply. We illustrate the control-theoretic relevance of these results by means of a few examples and design trade-offs.