Recent Network on Chip (NoC) design must take the thermal issue into consideration due to its great impact on the network performance and reliability, especially for 3D NoC. In this work, we design a virtual channel based fully adaptive routing algorithm for the runtime 3D NoC thermal-aware management. To improve the network throughput and latency, we use two virtual channels for each horizontal direction and design a routing function which can not only avoid deadlock and livelock, but also ensure high adaptivity and routability in the throttled network. For path selection, we design a strategy that takes priority to the distance, but also considers path diversity and traffic state. For throttling information collection, instead of transmitting the topology information of the whole network, we use a 12 bits register to reserve the router state for one hop away, which saves the hardware cost largely and decreases the network latency. In the experiments, we test our proposed routing algorithm in different states with different sizes, and the proposed algorithm shows better network latency and throughput with low power compared with traditional algorithms.

Channel state estimation-based backoff algorithms for channel access are being widely studied to solve wireless channel accessing and sharing problem especially in super dense wireless networks. In such algorithms, the precision of the channel state estimation determines the performance. How to make the estimation accurate in an efficient way to meet the system requirements is essential in designing the new channel access algorithms. In this paper, we first study the distribution and properties of inaccurate estimations using a novel biased estimation analysis model. We then propose an efficient backoff algorithm based on the theory of confidence interval estimation (BA-CIE), in which the minimum sample size is deduced to improve the contention window tuning efficiency, while a fault-tolerance interval structure is applied to reduce the inaccurate estimations so as to improve the contention window tuning accuracy. Our simulation results show that the throughput of our proposed BA-CIE algorithm can achieve 99% the theoretical maximum throughput of IEEE 802.11 networks, thanks to the improved contention window tuning performance.

Recovery of low-rank matrices has seen significant activity in many areas of science and engineering, motivated by theoretical results for exact reconstruction guarantees and interesting practical applications. Recently, numerous methods incorporated the nuclear norm to pursue the convexity of the optimization. However, this greatly restricts its capability and flexibility in dealing with many practical problems, where the singular values have clear physical meanings. This paper studies a generalized non-convex low-rank approximation, where the singular values are in lp-heuristic. Then specific results are derived for image restoration, including denoising and deblurring. Extensive experimental results on natural images demonstrate the improvement of the proposed method over the recent image restoration methods.