IEEE Home Search IEEE Shop Web Account Contact IEEE IEEE logo

IEEE Information Theory Society

Student Resources

 

Research Discussion and Lunch at ISIT 2006

 
     

Home

Student Resources

Info. Theory Society

Media Services

 

Committee Events

Future Events

Past Events

 

Volunteering

 

Getting Started

Info. Theory Textbook
Info. Theory Tutorials

Interference Channel
Comm. Networks
Cooperative Comm.
Cryptography
Multiuser Info. Theory
Network Coding

Biology & Info. Theory
Quantum Info. Theory
 

Doctoral Dissertations

 

Finding Your Career

Job Resources

Job Opportunities

Pool of Candidates

 

Useful Links

Writing Papers

Journals

Conferences

Resources

 

Committee Members

 

Thanks to all team leaders and students for their enthusiastic participation. The research round table, held for the first time at ISIT 2006, was a smashing success, judging by the size and participation. Here are the summaries of the discussions:
 

Table #1:            MIMO Broadcast Capacity and Space-Time Coding

                           Leader: Hanan Weingarten (Technion University)

                                        Jinsong Wu (Queen's University)
Papers:

  1. T. M. Cover, "Comments on Broadcast Channels"
    This is a short and very nice tutorial on the broadcast channel but does not contain what is known the MIMO broadcast channel.

  2. G. Caire and S. Shamai (Shitz), "On the Achievable Throughput of a Multiantenna Gaussian Broadcast Channel"
    This is the paper which launched the interest in MIMO broadcast channels and introduces Costa coding in the context of the broadcast channel.

  3. S. Vishwanath, N. Jindal and A. Goldsmith, "Duality, Achievable Rates, and Sum-Rate Capacity of Gaussian MIMO Broadcast Channels"
    This is a key paper on uplink-downlink duality and establishes the sum-capacity.

  4. H. Weingarten, Y. Steinberg and S. Shamai (Shitz), "The Capacity Region of the Gaussian Multiple Input Multiple Output Broadcast Channel", IEEE Transaction on Information Theory, Sept 2006
    This paper contains the converse for the capacity region of this channel and introduces the idea of channel enhancement.

Jinsong Wu: Quite a few people are interested in MIMO Broadcast (BC) channels. They exchanges their feelings on this topic. People also talked about their research interests and discussed specific questions, mainly on MIMO Broadcast, Scheduling.

Hanan Weingarten: The following only reflects part of discussion at our table: We discussed the differences between the proofs of MIMO broadcast channel capacity region given by myself and that given recently by Mehdi Mohseni. In addition we discussed briefly the problem of the capacity region of the MIMO BC with common information.

List of participants (Name, University, Research Interest):

Rui Zhang - Stanford - Dynamic resource allocations in MIMO-MAC, MIMO-BC, Recent results, Power region
Taesang Yoo - Stanford - MIMO-BC, Scheduling, Beamforming, Quantized feedback, Multiuser diversity
Cong Shen - UCLA - STC,Cross-layer resource allocation in MAC, feedback in 802.11n MIMO-OFDM
Mehdi Mohseni - Stanford - Gaussian MIMO BC, Resource allocation over fading MAC & BC
Jinsong Wu - Queen’s - STC, MIMO, MIMO-OFDM
Weingarten Hanan - Technion - IIT - MIMO BC
Tie Liu - UIUC - Entropy power inequalities
Camilla Hollanti - University of Turku, Finland

Kibeom Seong - Stanford - Cross layer resource allocation, BC & MAC
Mohammad A. Maddah-Ali - Waterloo - MIMO broadcast, Resource allocation, Multibase
 

Table #2:            Capacity with Feedback

                           Leader: Haim Permuter (Stanford University)

Papers:

  1. J Schalkwijk, T Kailath, "A coding scheme for additive noise channels with feedback I: No bandwidth constraint"

  2. L Ozarow, "The capacity of the white Gaussian multiple access channel with feedback"

  3. J. Massey, "Causality, Feedback and Directed Information"

  4. T. Cover, S. Pombra, "Gaussian Feedback Capacity"

  5. M.V. Burnashev "Data transmission over a discrete channel with feedback, random transmission time"

The round table discussion on feedback capacity had three main parts. In the first part each one introduced himself to the group. In the second one we did a review on feedback capacity and in the last part we have talked about research problems related to feedback capacity.

During the review on the feedback capacity several results among many that exists in the literature where mentioned. Here are only four of them:

Feedback does not increase capacity of Discrete Memoryless Channel (DMC):
This result was shown in 1956 by Shannon in [1]. In the paper Shanon distinguished between zero error capacity in which the error is identically zero to regular capacity where the error goes to zero as the block length of the code goes to infinity. Shannon showed that even though the feedback can increases the zero-error capacity it does not increase the regular capacity.

Schalkwijk-Kailath scheme
Schalkwijk and Kailath presented in 1966 a communication scheme [2] for memoryless Gaussian channel with feedback that shows that even though the capacity is not increased probability of error does significantly decrease compared to the nonfeedback setting the. More specifically, the probability of the decoding error decays doubly exponentially in the duration of communication, compared to the exponential decay for the nonfeedback setting.

Feedback capacity Gaussian channel with memory:
In 1989 Cover and Pombra [3] characterized the feedback capacity of channel with memory. By using the asymptotic equipartition theorem for non stationary Gaussian process they characterized the capacity of time-varying additive Gaussian noise. In addition, with the aid of certain matrix inequality they showed that feedback capacity can not increase by more then half a bit then the capacity without feedback and also can not increase more then twice the capacity without feedback. In this ISIT Young-Han Kim, who also participated in the round-table discussion, showed that this result can be used to find an explicit expression for the capacity of a channel with Auto-Regressive moving Average Gaussian noise of first order.

Directed Information:
In 1990 Massey [4] coined a new term called Directed information. Massey showed in [4] that directed information is always less or equal to the mutual information and in the case of no feedback they are equal. He also showed that directed information is an upper bound for the feedback capacity and conjectured that it is the right term for characterizing feedback capacity of channel with feedback. In this ISIT, Permuter, Weissman and Goldsmith proved that it is the right term for characterizing the capacity of channels with any time invariant deterministic feedback.

We would like to mention that there are many more important results related to feedback such as those that are related to multi-user settings like Broadcast, MAC relay and MIMO. Many of them were discussed during the third part of the discussion where many interesting research problem were suggested. In particular Brooke Shrader suggested the compound channel with feedback. Later on, we found out that the case, where each channel in the compound channel is memoryless, was considered by Wolfowitz in his book but the case that each channel in the compound channel has memory was not yet considered.

some references related to feedback capacity that were mentioned in the discussion:
[1] C. E. Shannon, "The Zero Error Capacity of a Noisy Channel." IEEE Trans. on Information Theory, Vol 2, pp. S8-S19, September 1956.
[2] J. P. M. Schalkwijk and T. Kailath, “A coding scheme for additive noise channels with feedback–I: No bandwidth constraint,” IEEE Trans. Inform. Theory, vol. IT-12,pp. 172–182, April 1966.
[3] T. Cover and S. Pombra, "Gaussian Feedback Capacity." IEEE Trans. on Information Theory, Vol. IT-35, No. 1, January 1989, pp. 37-43.
[4] J. L. Massey, Causality, feedback and directed information, Proceedings of the International Symposium on Information Theory and its Applications, Honolulu, HI, Nov. 27-30, 1990.

Table #3:            
Cooperation and Relaying

                            Leader: Chris Ng (Stanford University)

Papers:

  1. T. Cover and A. El Gamal, "Capacity Theorems for the Relay Channel"

We took turns to introduce ourselves and what we're working on. Most people are already working in the field of cooperative communications. A few were just starting in the topic.
We then discussed different cooperative protocols such as decode-and-forward and compress-and-forward. Different performance metrics were also discussed: ergodic capacity, outage capacity, diversity order etc. And we talked about the impact of different channel state information and power assumptions. Then the table broke out to smaller discussion groups with two to three people discussing the topics of their interests.
 

Table #4:            Network Coding

                           Leader: Niranjan Ratnakar (UIUC)

Papers:

  1. S.P Borade, "Network information flow: limits and achievability"

  2. R. Dougherty, C. Freiling, and K. Zeger "Unachievability of Network Coding Capacity"

The focus of this round table was on the interaction between information theory and network coding. The minutes of the meeting are as follows:

* The main idea of network coding was introduced. We are interested in communicating data across a network with multiple nodes (which represent sources, relays, and destinations). In the paradigm of network coding, data is encoded by all nodes in the network, as opposed to the traditional approach of encoding only at the source.

* Multicast (one source generates data that is demanded by multiple destinations) has been extensively studied by the network coding community.

* Multiple unicast (a unicast involves one source and one destination and by multiple unicast, we assume that there are parallel independent unicast sessions) is more difficult to analyze than multicast. Some of the results we discussed were the insufficiency of linear network coding (http://www.code.ucsd.edu/~zeger/publications/journals/DoFrZe05-IT-Insufficiency/DoFrZe05-IT-Insufficiency.pdf) and the recent result by Koetter (unpublished) where he considers a network of discrete memoryless channels (DMCs) and shows that an optimal mode of operating these networks is to perform channel coding on each DMC and then to perform network coding on this network of error free links.

* The above point suggests that the problem of computing capacity regions of networks of DMCs is the same as computing capacity regions of networks of error free links (which is a combinatoric problem). Was this a reason why comoputing the capacity regions of DMCs was an open problem for a long time?

* One of the questions raised during the discussion was whether capacity is the right measure to evaluate performance of networks? Note that capacity has been a successful measure for the point to
point channel.
 

Table #5:            Secrecy System Capacity
                           Leader: Yingbin Liang (Princeton University)

Papers:

  1. A. D. Wyner, "The Wire-Tap Channel"

  2. I. Csiszar and J. Korner "Broadcast Channels with Confidential Messages"

We first went over the Wyner's framework of information-theoretic security for the wire-tap channel, where a sender wishes to transmit confidential information to a legitimate receiver and to keep a wire-tapper as ignorant of this information as possible. A generalization of Wyner's wire-tap channel is the broadcast channel with confidential messages, and was studied by Csiszar and Korner. This model was also discussed.

The secrecy level of confidential information at the wire-tapper is measured by the equivocation rate, i.e., the entropy rate of confidential messages conditioned on channel outputs at the wire-tapper. This concept was introduced. The secrecy capacity was also discussed, which is the maximum rate that can be achieved with perfect secrecy, i.e., the wire-tapper is fully ignorant of the confidential message.

The important results on the secrecy capacity were summarized. These results include the secrecy capacity of the wire-tap channel and its Gaussian example, the broadcast channel with confidential messages, and the relay channel with confidential messages. The intuition behind the results were discussed.

The concept of common randomness was introduced. The secret key capacity that arises in this setting was discussed.

The security issues in fading wireless channels were finally discussed. One interesting problem in this setting is whether fading can help in achieving communication security. We also discussed what are important factors that affect secure communication in wireless channels.