arXiv subjects

Marcel Mross

Publications and source records attributed to Marcel Mross.

4 recordsLinked to original sources

arXiv2023

Second-order Rate Analysis of a Two-user Gaussian Interference Channel with Heterogeneous Blocklength Constraints

We consider a two-user Gaussian interference channel with heterogeneous blocklength constraints (HB-GIC), strong interference, and two private messages. We propose to apply the successive interference cancellation with early decoding, i.e., decoding a message with a number of received symbols less than the blocklength at the receiver. We determine the necessary number of received symbols to achieve successful decoding of the longer codeword that satisfies the input power constraints and target average error probability constraints. To attain the results, we investigate the dependence testing bound analysis over an independent and identically distributed (i.i.d.) Gaussian input. Besides, we derive the second-order achievable rate region of the considered HB-GIC. By numerical results based on the rate-profile approach, we compare the derived second-order rate region to the first-order one, which shows the rate back-off of the considered model due to the impact of finite blocklength.

cs.IT↗

arXiv2022

New Inner and Outer Bounds for Gaussian Broadcast Channels with Heterogeneous Blocklength Constraints

We investigate novel inner and outer bounds on the rate region of a 2-user Gaussian broadcast channel with finite, heterogeneous blocklength constraints (HB-GBC). In particular, we introduce a new, modified Sato-type outer bound that can be applied in the finite blocklength regime and which does not require the same marginal property. We then develop and analyze composite shell codes, which are suitable for the HB-GBC. Especially, to achieve a lower decoding latency for the user with a shorter blocklength constraint when successive interference cancellation is used, we derive the number of symbols needed to successfully early decode the other user's message. We numerically compare our derived outer bound to the best known achievable rate regions. Numerical results show that the new early decoding performance in terms of latency reduction is significantly improved compared to the state of the art, and it performs very close to the asymptotic limit.

cs.IT↗

arXiv2021

Gaussian Broadcast Channels under Heterogeneous Blocklength Constraints

Future wireless access networks aim to simultaneously support a large number of devices with heterogeneous service requirements, including data rates, error rates, and latencies. While achievable rate and capacity results exist for Gaussian broadcast channels in the asymptotic blocklength regime, the characterization of second-order achievable rate regions for heterogeneous blocklength constraints is not available. Therefore, we investigate a two-user Gaussian broadcast channel (GBC) with heterogeneous blocklength constraints, specified according to users' channel output signal-to-noise ratios (SNRs). We assume the user with higher output SNR has a shorter blocklength constraint. We show that with sufficiently large output SNR, the stronger user can perform the \textit{early decoding} (ED) technique to decode and subtract the interference via successive interference cancellation (SIC). To achieve it, we derive an explicit lower bound on the necessary number of received symbols for a successful ED, using an independent and identically distributed Gaussian input. A second-order rate of the weaker user who suffers from an SNR change due to the heterogeneous blocklength constraint, is also derived. Numerical results show that ED can outperform the hybrid non-orthogonal multiple access scheme when the stronger channel is sufficiently better than the weaker one. Under the considered setting, about 7-dB SNR gain can be achieved. These results shows that ED with SIC is a promising technique for the future wireless networks.

cs.IT↗

arXiv2021

Improving Online Performance Prediction for Semantic Segmentation

In this work we address the task of observing the performance of a semantic segmentation deep neural network (DNN) during online operation, i.e., during inference, which is of high importance in safety-critical applications such as autonomous driving. Here, many high-level decisions rely on such DNNs, which are usually evaluated offline, while their performance in online operation remains unknown. To solve this problem, we propose an improved online performance prediction scheme, building on a recently proposed concept of predicting the primary semantic segmentation task's performance. This can be achieved by evaluating the auxiliary task of monocular depth estimation with a measurement supplied by a LiDAR sensor and a subsequent regression to the semantic segmentation performance. In particular, we propose (i) sequential training methods for both tasks in a multi-task training setup, (ii) to share the encoder as well as parts of the decoder between both task's networks for improved efficiency, and (iii) a temporal statistics aggregation method, which significantly reduces the performance prediction error at the cost of a small algorithmic latency. Evaluation on the KITTI dataset shows that all three aspects improve the performance prediction compared to previous approaches.

cs.CV↗