| 1. |
- Devassy, Rahul, 1985, et al.
(författare)
-
Delay and Peak-Age Violation Probability in Short-Packet Transmissions
- 2018
-
Ingår i: IEEE International Symposium on Information Theory - Proceedings. - 2157-8095. ; 2018-June, s. 2471-2475
-
Konferensbidrag (refereegranskat)abstract
- This paper investigates the distribution of delay and peak age of information in a communication system where packets, generated according to an independent and identically distributed Bernoulli process, are placed in a single-server queue with first-come first-served discipline and transmitted over an additive white Gaussian noise (AWGN) channel. When a packet is correctly decoded, the sender receives an instantaneous error-free positive acknowledgment, upon which it removes the packet from the buffer. In the case of negative acknowledgment, the packet is retransmitted. By leveraging finite-blocklength results for the AWGN channel, we characterize the delay violation and the peak-age violation probability without resorting to approximations based on large deviation theory as in previous literature. Our analysis reveals that there exists an optimum blocklength that minimizes the delay violation and the peak-age violation probabilities. We also show that one can find two blocklength values that result in very similar average delay but significantly different delay violation probabilities. This highlights the importance of focusing on violation probabilities rather than on averages.
|
|
| 2. |
- Devassy, Rahul, 1985, et al.
(författare)
-
Reliable Transmission of Short Packets Through Queues and Noisy Channels Under Latency and Peak-Age Violation Guarantees
- 2019
-
Ingår i: IEEE Journal on Selected Areas in Communications. - 0733-8716 .- 1558-0008. ; 37:4, s. 721-734
-
Tidskriftsartikel (refereegranskat)abstract
- This paper investigates the probability that the delay and the peak-age of information exceed a desired threshold in a point-to-point communication system with short information packets. The packets are generated according to a stationary memoryless Bernoulli process, placed in a single-server queue and then transmitted over a wireless channel. A variable-length stop-feedback coding scheme - a general strategy that encompasses simple automatic repetition request (ARQ) and more sophisticated hybrid ARQ techniques as special cases - is used by the transmitter to convey the information packets to the receiver. By leveraging finite-blocklength results, the delay violation and the peak-age violation probabilities are characterized without resorting to approximations based on larg-deviation theory as in previous literature. Numerical results illuminate the dependence of delay and peak-age violation probability on system parameters such as the frame size and the undetected error probability, and on the chosen packet-management policy. The guidelines provided by our analysis are particularly useful for the design of low-latency ultra-reliable communication systems.
|
|
| 3. |
- Jose, Sharu Theresa, et al.
(författare)
-
Transfer Meta-Learning: Information-Theoretic Bounds and Information Meta-Risk Minimization
- 2022
-
Ingår i: IEEE Transactions on Information Theory. - 0018-9448 .- 1557-9654. ; 68:1, s. 474-501
-
Tidskriftsartikel (refereegranskat)abstract
- Meta-learning automatically infers an inductive bias by observing data from a number of related tasks. The inductive bias is encoded by hyperparameters that determine aspects of the model class or training algorithm, such as initialization or learning rate. Meta-learning assumes that the learning tasks belong to a task environment, and that tasks are drawn from the same task environment both during meta-training and meta-testing. This, however, may not hold true in practice. In this paper, we introduce the problem of transfer meta-learning, in which tasks are drawn from a target task environment during meta-testing that may differ from the source task environment observed during meta-training. Novel information-theoretic upper bounds are obtained on the transfer meta-generalization gap, which measures the difference between the meta-training loss, available at the meta-learner, and the average loss on meta-test data from a new, randomly selected, task in the target task environment. The first bound, on the average transfer meta-generalization gap, captures the meta-environment shift between source and target task environments via the KL divergence between source and target data distributions. The second, PAC-Bayesian bound, and the third, single-draw bound, account for this shift via the log-likelihood ratio between source and target task distributions. Furthermore, two transfer meta-learning solutions are introduced. For the first, termed Empirical Meta-Risk Minimization (EMRM), we derive bounds on the average optimality gap. The second, referred to as Information Meta-Risk Minimization (IMRM), is obtained by minimizing the PAC-Bayesian bound. IMRM is shown via experiments to potentially outperform EMRM.
|
|
| 4. |
- Popovski, Petar, et al.
(författare)
-
5G wireless network slicing for eMBB, URLLC, and mMTC: A communication-theoretic view
- 2018
-
Ingår i: IEEE Access. - 2169-3536 .- 2169-3536. ; 6, s. 55765-55779
-
Tidskriftsartikel (refereegranskat)abstract
- The grand objective of 5G wireless technology is to support three generic services with vastly heterogeneous requirements: enhanced mobile broadband (eMBB), massive machine-type communications (mMTCs), and ultra-reliable low-latency communications (URLLCs). Service heterogeneity can be accommodated by network slicing, through which each service is allocated resources to provide performance guarantees and isolation from the other services. Slicing of the radio access network (RAN) is typically done by means of orthogonal resource allocation among the services. This paper studies the potential advantages of allowing for non-orthogonal sharing of RAN resources in uplink communications from a set of eMBB, mMTC, and URLLC devices to a common base station. The approach is referred to as heterogeneous non-orthogonal multiple access (H-NOMA), in contrast to the conventional NOMA techniques that involve users with homogeneous requirements and hence can be investigated through a standard multiple access channel. The study devises a communication-theoretic model that accounts for the heterogeneous requirements and characteristics of the three services. The concept of reliability diversity is introduced as a design principle that leverages the different reliability requirements across the services in order to ensure performance guarantees with non-orthogonal RAN slicing. This paper reveals that H-NOMA can lead, in some regimes, to significant gains in terms of performance tradeoffs among the three generic services as compared to orthogonal slicing.
|
|
| 5. |
- Rezazadeh, Arezou, 1987, et al.
(författare)
-
Conditional Mutual Information-Based Generalization Bound for Meta Learning
- 2021
-
Ingår i: IEEE International Symposium on Information Theory - Proceedings. - 2157-8095. ; 2021-July, s. 1176-1181
-
Konferensbidrag (refereegranskat)abstract
- Meta-learning optimizes an inductive bias—typically in the form of the hyperparameters of a base-learning algorithm—by observing data from a finite number of related tasks. This paper presents an information-theoretic bound on the generalization performance of any given meta-learner, which builds on the conditional mutual information (CMI) framework of Steinke and Zakynthinou (2020). In the proposed extension to meta-learning, the CMI bound involves a training meta-supersample obtained by first sampling 2N independent tasks from the task environment, and then drawing 2M independent training samples for each sampled task. The meta-training data fed to the meta-learner is modelled as being obtained by randomly selecting N tasks from the available 2N tasks and M training samples per task from the available 2M training samples per task. The resulting bound is explicit in two CMI terms, which measure the information that the meta-learner output and the base-learner output provide about which training data are selected, given the entire meta-supersample. Finally, we present a numerical example that illustrates the merits of the proposed bound in comparison to prior information-theoretic bounds for meta-learning
|
|
