Beijing University of Posts and Telecommunications

About: Beijing University of Posts and Telecommunications is a education organization based out in Beijing, Beijing, China. It is known for research contribution in the topics: MIMO & Quality of service. The organization has 39576 authors who have published 41525 publications receiving 403759 citations. The organization is also known as: BUPT.

Adaptive Dilated Network With Self-Correction Supervision for Counting

Abstract: The counting problem aims to estimate the number of objects in images. Due to large scale variation and labeling deviations, it remains a challenging task. The static density map supervised learning framework is widely used in existing methods, which uses the Gaussian kernel to generate a density map as the learning target and utilizes the Euclidean distance to optimize the model. However, the framework is intolerable to the labeling deviations and can not reflect the scale variation. In this paper, we propose an adaptive dilated convolution and a novel supervised learning framework named self-correction (SC) supervision. In the supervision level, the SC supervision utilizes the outputs of the model to iteratively correct the annotations and employs the SC loss to simultaneously optimize the model from both the whole and the individuals. In the feature level, the proposed adaptive dilated convolution predicts a continuous value as the specific dilation rate for each location, which adapts the scale variation better than a discrete and static dilation rate. Extensive experiments illustrate that our approach has achieved a consistent improvement on four challenging benchmarks. Especially, our approach achieves better performance than the state-of-the-art methods on all benchmark datasets.

Fine-Grained Visual Classification via Progressive Multi-granularity Training of Jigsaw Patches

Abstract: Fine-grained visual classification (FGVC) is much more challenging than traditional classification tasks due to the inherently subtle intra-class object variations. Recent works are mainly part-driven (either explicitly or implicitly), with the assumption that fine-grained information naturally rests within the parts. In this paper, we take a different stance, and show that part operations are not strictly necessary – the key lies with encouraging the network to learn at different granularities and progressively fusing multi-granularity features together. In particular, we propose: (i) a progressive training strategy that effectively fuses features from different granularities, and (ii) a random jigsaw patch generator that encourages the network to learn features at specific granularities. We evaluate on several standard FGVC benchmark datasets, and show the proposed method consistently outperforms existing alternatives or delivers competitive results. The code is available at https://github.com/PRIS-CV/PMG-Progressive-Multi-Granularity-Training.

Resource Allocation for Ultra-Dense Networks: A Survey, Some Research Issues and Challenges

Abstract: Driven by the explosive data traffic and new quality of service requirement of mobile users, the communication industry has been experiencing a new evolution by means of network infrastructure densification. With the increase of the density as well as the variety of access points (APs), the network benefits from proximal transmissions and increased spatial reuse of system resources, thus introducing a new paradigm named ultra-dense networks (UDNs). Since the limited available resources are shared by ubiquitous APs in UDNs, the demand for efficient resource allocation schemes becomes even more compelling. However, the large scale of UDNs impedes the exploration of effective resource allocation approaches particularly on the computational complexity and significance overhead or feedback. In this paper, we provide a survey-style introduction to resource allocation approaches in UDNs. Specifically, we first present some common scenarios of UDNs with the relevant special issues. Second, we provide a taxonomy to classify the resource allocation methods in the existing literatures. Then, to alleviate the main difficulties of UDNs, some prevailing and feasible solutions are elaborated. Next, we present some emerging technologies thriving UDNs with special RA features discussed. Additionally, the challenges and open research directions are outlined in this field.

Energy-saving virtual machine placement in cloud data centers

Abstract: In cloud data centers, different mapping relationships between virtual machines (VMs) and physical machines (PMs) cause different resource utilization, therefore, how to place VMs on PMs to improve resource utilization and reduce energy consumption is one of the major concerns for cloud providers. The existing VM placement schemes are to optimize physical server resources utilization or network resources utilization, but few of them focuses on optimizing multiple resources utilization simultaneously. To address the issue, this paper proposes a VM placement scheme meeting multiple resource constraints, such as the physical server size (CPU, memory, storage, bandwidth, etc.) and network link capacity to improve resource utilization and reduce both the number of active physical servers and network elements so as to finally reduce energy consumption. Since VM placement problem is abstracted as a combination of bin packing problem and quadratic assignment problem, which is also known as a classic combinatorial optimization and NP-hard problem, we design a novel greedy algorithm by combining minimum cut with the best-fit, and the simulations show that our solution achieves better results.

Self-supervised Heterogeneous Graph Neural Network with Co-contrastive Learning

Abstract: Heterogeneous graph neural networks (HGNNs) as an emerging technique have shown superior capacity of dealing with heterogeneous information network (HIN). However, most HGNNs follow a semi-supervised learning manner, which notably limits their wide use in reality since labels are usually scarce in real applications. Recently, contrastive learning, a self-supervised method, becomes one of the most exciting learning paradigms and shows great potential when there are no labels. In this paper, we study the problem of self-supervised HGNNs and propose a novel co-contrastive learning mechanism for HGNNs, named HeCo. Different from traditional contrastive learning which only focuses on contrasting positive and negative samples, HeCo employs cross-view contrastive mechanism. Specifically, two views of a HIN (network schema and meta-path views) are proposed to learn node embeddings, so as to capture both of local and high-order structures simultaneously. Then the cross-view contrastive learning, as well as a view mask mechanism, is proposed, which is able to extract the positive and negative embeddings from two views. This enables the two views to collaboratively supervise each other and finally learn high-level node embeddings. Moreover, two extensions of HeCo are designed to generate harder negative samples with high quality, which further boosts the performance of HeCo. Extensive experiments conducted on a variety of real-world networks show the superior performance of the proposed methods over the state-of-the-arts.