科研成果 by Type: Conference Paper

Mobile edge computing (MEC) has been an effective paradigm to support real-time computation-intensive vehicular applications. However, due to highly dynamic vehicular topology, these existing centralized-based or distributed-based scheduling algorithms requiring high communication overhead, are not suitable for task offloading in vehicular networks. Therefore, we investigate a novel service scenario of MEC-based vehicular crowdsourcing, where each MEC server is an independent agent and responsible for making scheduling of processing traffic data sensed by crowdsourcing vehicles. On this basis, we formulate a data-driven task offloading problem by jointly optimizing offloading decision and bandwidth/computation resource allocation, and renting cost of heterogeneous servers, such as powerful vehicles, MEC servers and cloud, which is a mixed-integer programming problem and NP-hard. To reduce high time-complexity, we propose the solution in two stages. First, we design an asynchronous deep Q-learning to determine offloading decision, which achieves fast convergence by training the local DQN model at each agent in parallel and uploading for global model update asynchronously. Second, we decompose the remaining resource allocation problem into several independent subproblems and derive optimal analytic formula based on convex theory. Lastly, we build a simulation model and conduct comprehensive simulation, which demonstrates the superiority of the proposed algorithm.

This paper examines a Sinopean victory list of the boxer Marcus Iutius Marcianus Rufus (French 2004: 76-77 no. 105) and the implications of counting the number of victories he won. Inscribed and set up by the Sinopean boule, the list represents an official recognition of the athlete's successful boxing career, which not only included victories in the four periodoi of mainland Greece, but also the Capitoline and Neapolitan games in Italy. The text has been studied by Theodoré Reinach (1916), George Bean (1953), and David French (2004), and resulting in different ways to count Rufus' victories.The three epigraphists encountered several issues with counting Rufus' victories. How to differentiate between a Bithynian koinon event from a metropolitan event held by Nicaea and Nicomedia is one issue, and whether to count the half-talent victories with the iselastic victories so to fit an ideal number of total victories that Rufus won is another, with the three epigraphists producing different solutions. Perhaps more perplexing of all, however, is how to interpret the Greek letters ΡΝ placed at the end of the victory list. Reinach interpreted them as the remaining letters of ἀνδριατί or "jeux mineurs" (Reinach 1916: 358). Bean and French saw them as Greek numerals, indicating the total tally of all listed victories. While the total tally seems a convincing interpretation on formulaic grounds, the arithmetic does not add up. On the one hand, tabulation indicates that Bean's count of total victories yields 159, with 110 half-talent victories and 49 iselastic victories. He reconciled the number by claiming to have seen signs of reinscribing in the squeeze, and suggested that Rufus initially won 101 half-talent victories, only to have achieved 110 at a later time, upon which occasion an update was applied to his monument (Bean 1953: 176). On the other hand, while French counted the half-talent victories as 110, and his total number of iselastic victories amount to 48, he still maintained that ΡΝ stands for "(In all) 150 victories," leaving the arithmetic issue open for further examination (French 2004: 77).This paper surveys other victory lists to study how koinon and metropolitan victories were differentiated and counted, and how chronographic features were positioned and identified. This paper also proposes to disassociate the number 150 from the total count of victories, and reconsider what was signified by this number. One possibility is the era: the 150th year of the era of Sinope. It has been demonstrated that Sinopean coinage during the imperial period used first the colonial era from 45 BCE, then the so-called Lucullan era of 70 BCE (Leschhorn 1993: 161-162). While era-based chronography is not found on extant imperial period inscriptions from Sinope, Rufus' victory list may be the first surviving example.

Deep Spiking Neural Networks (SNNs) present optimization difficulties for gradient-based approaches due to discrete binary activation and complex spatial-temporal dynamics. Considering the huge success of ResNet in deep learning, it would be natural to train deep SNNs with residual learning. Previous Spiking ResNet mimics the standard residual block in ANNs and simply replaces ReLU activation layers with spiking neurons, which suffers the degradation problem and can hardly implement residual learning. In this paper, we propose the spike-element-wise (SEW) ResNet to realize residual learning in deep SNNs. We prove that the SEW ResNet can easily implement identity mapping and overcome the vanishing/exploding gradient problems of Spiking ResNet. We evaluate our SEW ResNet on ImageNet, DVS Gesture, and CIFAR10-DVS datasets, and show that SEW ResNet outperforms the state-of-the-art directly trained SNNs in both accuracy and time-steps. Moreover, SEW ResNet can achieve higher performance by simply adding more layers, providing a simple method to train deep SNNs. To our best knowledge, this is the first time that directly training deep SNNs with more than 100 layers becomes possible. Our codes are available at https://github.com/fangwei123456/Spike-Element-Wise-ResNet.