The distributed remote source coding (the so-called CEO) problem is studied in the case where the underlying source, not necessarily Gaussian, has finite differential entropy and the observation noise is Gaussian. The main result is a new lower bound for the sum-rate-distortion function under arbitrary distortion measures. When specialized to the case of mean-squared error, it is shown that the bound exactly mirrors a corresponding upper bound, except that the upper bound has the source power (variance), whereas the lower bound has the source entropy power. Bounds exhibiting this pleasing duality of power and entropy power have been well known for direct and centralized source coding since Shannon's work. While the bounds hold generally, their value is most pronounced when interpreted as a function of the number of agents in the CEO problem.

This paper explores the problem of feedback coding for a channel whose output is simultaneously used for two purposes: it is decoded to establish reliable communication, and it is used to control a dynamical system. In general, there is a tradeoff between the rate of communication and the accuracy of control. An intuitive communication and control strategy is analyzed and shown to be optimal in several cases of interest. Using methods from stochastic control, a corresponding upper bound is derived on the rate for a given cost, and this bound can be applied to find the capacity for an interesting class of channels. Under certain regularity conditions, the capacity has a simple characterization, and a series of examples is provided to demonstrate how it can be calculated.

Alice and Bob want to share a secret key and to communicate an independent message, both of which they desire to be kept secret from an eavesdropper Eve. This problem of secret communication and secret-key generation when two resources are available-correlated sources at Alice, Bob, and Eve, and a noisy broadcast channel from Alice to Bob and Eve which is independent of the sources is studied. The goal is to characterize the fundamental tradeoff between the rates of the secret message and secret key. An achievable solution and proof of its optimality for the parallel channels and sources case when each subchannel and source component satisfies a degradation order (either in favor of the legitimate receiver or the eavesdropper) is presented. This includes the case of jointly Gaussian sources and an additive Gaussian channel, for which the secrecy region is evaluated.

A fundamental problem in dynamic frequency reuse is that the cognitive radio is ignorant of the amount of interference it inflicts on the primary license holder. Policies that attempt to limit interference without the active participation of the primary are thus difficult to implement. However, many wireless systems use flow control feedback such as ARQs. By listening to these control signals, a cognitive radio can obtain indirect information about the interference it generates and thus behave in an acceptable manner. This paper introduces an information-theoretic model of this basic observation and develops and analyzes algorithms that can exploit it. In particular, a simple generic strategy is proposed where the cognitive radio monitors the primary's effective packet rate and only transmits when that rate is above a threshold. The strategy is shown to have important universality properties with respect to unknown time-varying interference characteristics as well as favorable delay properties.

The utility of limited feedback for coding over an individual sequence of discrete memoryless channels is investigated. This study complements recent results showing how limited or noisy feedback can boost the reliability of communication. A strategy with fixed input distributionPis given that asymptotically achieves rates arbitrarily close to the mutual information induced byPand the state-averaged channel. When the capacity-achieving input distribution is the same over all channel states, this achieves rates at least as large as the capacity of the state-averaged channel, sometimes called the empirical capacity.

A generalization of the additive Gaussian two-way channel toMusers is considered. Such channels contain implicit feedback in the sense that the channel output signals observed by the different encoders are correlated. While the benefits of feedback are shown to be negligible at high SNR, for moderate SNR, feedback can play a significant role in boosting the sum-rate performance. To highlight this potential gain, the special case of theM-user multiway channel with a common output is considered. By taking insights from Kramerpsilas Fourier MEC, a feedback strategy is introduced and shown to strictly dominate the performance of a pre-log optimal non-feedback strategy. Furthermore, an upper bound is derived to show this feedback strategy achieves the sum-rate capacity beyond a certain SNR threshold. Under per-symbol power constraints, this upper bound can be tightened to show the feedback strategy is sum-rate optimal for all SNR values.

Alice and Bob want to share a secret key and to communicate an independent message, both of which they desire to be kept secret from an eavesdropper Eve. We study this problem of secret communication and secret key generation when two resources are available — correlated sources at Alice, Bob, and Eve, and a noisy broadcast channel from Alice to Bob and Eve. No other resource, in particular, no other channel is available. We are interested in characterizing the fundamental trade-off between the rates of the secret message and secret key. We present an achievable solution based on a separation architecture and prove its optimality under three settings: when Eve’s source and channel are degraded versions of Bob’s, and either Bob’s source or channel is by itself useless in generating a secret key.

We study a problem motivated by cognitive radio in which the primary is a packet system that employs ARQ feedback. A secondary system is allowed to transmit in the same frequency band provided it ensures that the primary attains a specified target rate. That is, the secondary has a certain "interference budget". The crux of the problem is that the secondary does not know how much interference it creates on the primary and therefore is ignorant of its interference budget. Absent this knowledge, we propose a scheme in which the secondary eavesdrops on the primary's ARQ and uses this knowledge to stay within its interference budget. Under certain assumptions, we show there exists an optimal rate-interference budget (RIB) tradeoff. We compare how far fixed strategies are from this RIB function as we vary the interference budget. Further, we exhibit a strategy that is optimal beyond a threshold interference budget and within 1 bit per primary packet elsewhere.

Recently, Shayevitz and Feeler introduced an individual sequence formulation of channel coding under model uncertainty and an elegant coding strategy that adapts Horstein's scheme to this setting to achieve the empirical capacity of the channel. Their scheme requires both full-rate output feedback and common randomness. We present a strategy in the style of Hybrid ARQ that requires no output feedback by using common randomness and zero-rate active feedback. This strategy still asymptotically achieves the empirical capacity.

An efficient distributed source coding system with two encoders and dependent data streams must remove two kinds of redundancy: redundancy in each stream and between the two streams. The striking result of Slepian and Wolf showed that the latter can be eliminated even if each encoder only observes one of the source streams. The coding technique that permits to achieve this is often referred to as "binning." In large source networks, binning can result in considerable savings in terms of encoding rate. The focus of this paper is on the scaling-law behavior, i.e., the characteristic performance in the limit as the source network size tends to infinity. For paradigmatic network topologies, we analyze the rate savings through binning, and we show that in some cases of interest, binning is scaling-law irrelevant.

In the CEO problem, introduced by Berger et al, IEEE Trans. Info. Theory, 1996, a CEO is interested in a source that cannot be observed directly. M agents observe independently noisy versions of the source and, without collaborating, must encode these across noiseless rate-constrained channels to the CEO. The quadratic AWGN CEO problem refers to the class of CEO problems for which the agents view the source through additive white Gaussian noise, and the distortion is squared error. This paper discusses two upper bounds to the CEO sum-rate distortion function for this class of problems. The first follows from elementary arguments. It permits two conclusions. First, the worst case is when the underlying source is Gaussian (for fixed variance). Second, there are source distributions that lead to a significantly better behavior. The second upper bound follows from a new bound on the rate loss between the CEO and the remote rate-distortion function. For certain source distributions and certain ranges of distortion, this bound is better than the first