Histopathological images provide the medical evidences to help the disease diagnosis. However, pathologists are not always available or are overloaded by work. Moreover, the variations of pathological images with respect to different organs, cell sizes and magnification factors lead to the difficulty of developing a general method to solve the histopathological image classification problems. To address these issues, we propose a novel cross-scale fusion (CSF) transformer which consists of the multiple field-of-view patch embedding module, the transformer encoders and the cross-fusion modules. Based on the proposed modules, the CSF transformer can effectively integrate patch embeddings of different field-of-views to learn cross-scale contextual correlations, which represent tissues and cells of different sizes and magnification factors, with less memory usage and computation compared with the state-of-the-art transformers. To verify the generalization ability of the CSF transformer, experiments are performed on four public datasets of different organs and magnification factors. The CSF transformer outperforms the state-of-the-art task specific methods, convolutional neural network-based methods and transformer-based methods. The source code will be available in our GitHub https://github.com/nchucvml/CSFT.

This paper proposes a network-aware resource management scheme that improves the quality of experience (QoE) for adaptive video streaming in CDNs and Information-Centric Networks (ICN) in general, and Dynamic Adaptive Streaming over HTTP (DASH) in particular. By utilizing the DASH manifest, the network (by way of a logically centralized controller) computes the available link resources and schedules the chunk dissemination to edge caches ahead of the end-user's requests. Our approach is optimized for multi-rate DASH videos. We implemented our resource management scheme, and demonstrated that in the scenario when network conditions evolve quickly, our approach can maintain smooth high quality playback. We show on actual video server data and in our own simulation environment that a significant reduction in peak bandwidth of 20% can be achieved using our approach.

While geo-routing is a promising routing algorithm in urban vehicular ad-hoc networks (VANETs), there is still much work to be done for it to become truly usable for such environments. One of the biggest obstacles to this goal is the intermittent nature of VANETs due to mobility. Traditional geo-routing algorithms do not perform well in these conditions because they drop packets whenever they cannot find an immediate forwarder for the packet. In this paper, we propose RobustGeo, a routing protocol that combines the simplicity and efficiency of the greedy forwarding technique in geo-routing algorithms, with the robustness of delay tolerant networks in the face of disruptions in network connectivity. When there exists a good connection between the source and destination, RobustGeo can route the packet like the traditional geo-routing algorithms, and when the network faces disruptions, RobustGeo relies upon vehicle mobility and packet replication to explore multiple georoute paths and quickly recover the packet back to greedy forwarding. We show that for a highly intermittent scenario, RobustGeo has a delivery ratio of over 20% (compared to a pure geo-routing protocol's delivery ratio of 0), and it reduces delay by over 40% as compared to a more pure delay-tolerant routing solution.

Abstract In this paper, we discuss scalable content‐oriented routing that enables storing, sharing, and searching data totally within the urban VANET. We introduce a scalable proactive content discovery scheme, hierarchical bloom‐filter routing (HBFR), to tackle mobility, large population, and rich content challenges of VANETs. Our simulation results show that HBFR achieves low latency and high content availability with relatively low overhead in practical VANET scenarios. In particular, HBFR improves the response time by up to 45% compared with the conventional Content Centric Networking (CCN) and reduces the traffic by up to 85% compared with timer‐based intelligent content search. HBFR suits the popular private data services the best and also achieves low response time comparable with that of flooding‐based content discovery when applying to popular public data services.Copyright © 2014 John Wiley & Sons, Ltd.

Recently, Information Centric Networking (ICN) has attracted much attention also for mobiles. Unlike host-based communication models, ICN promotes data names as the first-class citizen in the network. However, the current ICN name-based routing requires Interests be routed by name to the nearest replica, implying the Interests are flooded in VANET. This introduces large overhead and consequently degrades wireless network performance. In order to maintain the efficiency of ICN implementation in VANET, we propose an opportunistic geo-inspired content based routing method. Our method utilizes the last encounter information of each node to infer the locations of content holders. With this information, the Interests can be geo-routed instead of being flooded to reduce the congestion level of the entire network. The simulation results show that our proposed method reduces the scope of flooding to less than two hops and improves retrieval rate by 1.42 times over flooding-based methods.

In this paper, we discuss scalable content-oriented routing that enables storing, sharing and searching data totally within the VANET. We introduce a proactive content discovery scheme, Hierarchical Bloom-Filter Routing (HBFR), to tackle mobility, large population and rich content challenges of VANETs. HBFR is compared to the popular ICN reactive content discovery scheme in practical VANET scenarios. The results show that HBFR suits non-sharable data services, while reactive ICN inspired content discovery works well with popular sharable data. We suggest a hybrid approach that adaptively utilizes proactive and reactive schemes for time-sensitive data in ICN VANET.

First responders at the edge of the network rely on situation awareness updates to arrive in a timely matter, even when the fixed infrastructure is unavailable. The technical advancements of the commercial mobile phones make them capable of supporting such requirements under very disruptive network conditions. In this demo, we present a network architecture that exploits partial caches by utilizing network coding to deliver large files (e.g. images) to first responders. The architecture is based on a content centric network platform called ICEMAN(Information CEntric Mobile Ad-hoc Networking) and runs on Android phones. We demonstrate our system in a file dissemination scenario in a CORE/EMANE network emulation. We measure file delivery ratio, latency, and network overhead and report significant improvements that network coding achieves over fragmentation.

TCP over Mobile Ad-hoc Networks (MANETs) is challenging due to frequent route breaks, high random errors, and DATA-ACK interference. Network coded multipath approaches have been shown in several previous studies to be an effective transmission paradigms in disruptive networks. However, most of the previous studies either have no adaptive redundancy control or rely on theoretical models that require knowledge at all relays of the entire network state. In this paper, we propose a network coded multipath scheme for conventional TCPCodeMP that adapts to frequent link changes in MANET and requires no explicit control messages. The scheme exploits multiple-path redundancy and maintains total transparency to transport layer protocols. The proposed coding scheme is based on three components: (1) random linear coding scheme with adjustable redundancy, (2) multipath routing, (3) ACK Piggy coding. Simulation results show that in a 3-hop static scenario, the proposed multipath scheme improves TCP flows goodput by 70% compared to our previous single-path scheme. In an extreme MANET scenario where two TCP sessions co-exists and nodes are moving as fast as 25 m/s with up to 40% packet error rate (an environment in which regular TCP collapses completely), CodeMP achieves at least 700Kbps aggregate TCP goodput, with a Jain's fairness index of 0.99.

Network coding has attracted much attention in recent years due to the capability of transmission reduction and throughput improvement. In this paper, a redundancy-based delivery mechanism is proposed. This delivery mechanism utilizes network coding and linear combination on original packets to regenerate packets, which are called coded packets. Upon receiving a minimum number of coded packets, the destination can decode and get all original packets. Through theoretical analysis on delivery rate and overhead, we further propose an utility function to determine a proper redundancy degree and observe the suitable range of packet error rates to implement our delivery mechanism. Furthermore, the proposed delivery mechanism is applied to link adaptation. The simulation results show that by utilizing our proposed delivery, the operating SNR range of each PHY mode is lowered. In other words, when channel condition gets worse, one mobile device can still operate in the same PHY mode by utilizing the proposed delivery mechanism.

IEEE 802.16 Broadband Wireless Access system (BWA) is a promising emerging technology. In the IEEE 802.16 standard, multiple data transmission rates are defined by employing different modulation and channel coding schemes. How to select a proper mode affects the system performance significantly. In this paper, an efficient link rate adaptation algorithm, named QoS-aware Link Rate Adaptation (Q-LRA), is proposed. Q-LRA takes both QoS demand and channel condition into consideration to determine the best mode. The QoS demand of a connection is represented as packet error rate (PER). Based on the specified PER, the BS calculates the required signal-to-noise ratio (SNR) threshold of each mode. By comparing the measured SNR and all modes' SNR thresholds, the BS suggests the SS the best mode. The performance of Q-LRA is evaluated by simulations. The simulation results show that Q-LRA improves system throughput and guarantees each connection's QoS demand.