Performing better text translation by integrating auxiliary inputs from visual information has gained widespread attention in recent years. While existing methods outperform the text-only translation models, the step-by-step generative style reduces the inference speed, which limits their applicability in real-world scenarios. In this paper, we propose the non-autoregressive language model (NA-LM) for multimodal machine translation. With NA-LM, we develop a Non-Autoregressive Multimodal Transformer (NA-MMT), which accelerates the generative translation via a parallel multimodal decoder. To retain the translation performance, we improve the NA-MMT in twofold: 1) We preprocess the image into a refined sequence of visual entities with length encoding to reduce irrelevant information; 2) We design cross fertility and cross-modal gate attention for multimodal decoder to enhance the generative quality. Experiments on Multi30k datasets show the NA-MMT can generate high-quality translation with over 11× speedup than the baselines, which is strongly competitive.

Non-autoregressive Transformers (NATs) have garnered significant attention due to their efficient decoding compared to autoregressive methods. However, existing conditional dependency modeling schemes based on masked language modeling introduce a training-inference gap in NATs. For instance, while NATs sample target words during training to enhance input, this condition cannot be met during inference, and simply annealing the sampling rate to zero during training leads to model performance degradation. We demonstrate that this training-inference gap prevents NATs from fully realizing their potential. To address this, we propose an adaptive end-to-end quantization alignment training framework, which introduces a semantic consistency space to adaptively align NAT training, eliminating the need for target information and thereby bridging the training-inference gap.Experimental results demonstrate that our method outperforms most existing fully NAT models, delivering performance on par with Autoregressive Transformer (AT) while being 17.0 times more efficient in inference.

With the benefit of virtualization technology and an ascending trend towards more cloud applications, virtual cluster (VC) allocation concerning different requirements of application deployment is vital to cloud datacenters and edge networks. Most of the state-of-the-art approaches for VC allocation problems are probability-based heuristics, and none of them absolutely outperforms any other approach in any case. Thus, it is necessary to collect and select effective and efficient approaches to jointly solve diverse and customizable VC allocation problems. This paper presents VCA-Solver, a VC allocation solver that can draw on collectively feasible schemes from multiple effective approaches. We describe the architecture of VCA-Solver and present a preliminary evaluation with a prototype implementation.

Mobile edge networks enhance cloud applications by offering reduced delay for mobile terminals. However, due to the cumulative increase of mobile terminals and applications, edge networks fail to accommodate all applications and are difficult to guarantee the end-to-end delay demand, which is an important metric of QoS. One potential solution to address this challenge is to place virtual clusters (VCs) in mobile micro-clouds (MMCs) to enhance the capabilities of edge networks and the cloud. Therefore, in this paper, we build the optimization model of VC placement in edges for maximizing the benefit function, with constraints of 1). the satisfaction for end-to-end delay demand of each application; 2). the resource capacity of each MMC no more than the whole demands of those VCs placed in. We use Stochastic Network Calculus (SNC) to build the delay model more accurately, considering stochastic arrivals of flows and stochastic services of service nodes. Moreover, we use Martingale Theory to enhance the solvability of the SNC-based delay model, achieving tighter delay bound reliability.

In this paper, we propose a system framework of Pub/Sub-based multimodal data transmission to mobile terminals. For various sensors, we classify the data in different topics, but transmit them in the uniform data channel, with different transmission mechanisms. In this case, mobile terminals receive data via different mechanisms, the push notification and request-response connection. The proposed framework makes the data transmission more efficient and flexible.

Weakly supervised object localization (WSOL) aims at pre-dicting the location of objects with image-level labels. Fine-grained WSOL task has its characteristic challenge compared with generic object localization. The structural information of the objects in fine-grained benchmark has little relevance to class. Previous WSOL works mainly focus on learning the most class discriminative parts recursively, leading to se-rious structural feature missing issue. In this paper, we pro-pose a self-guided network (SGN) which consists of two branch deep classification networks. It adopts a coarse-to-fine strategy to detect the structural information of the ob-ject. First, we devise a self-adaptive method (SAM) to de-tect the most body structure of the object by directly leveraging the feature recognition ability of the first classifier. Then, an object structure generation (OSG) method is proposed in the fine localization phase. OSG helps the second classifier to learn the boundary feature of the object with less back-ground noise. Extensive experiments on four well-known fine-grained benchmarks, including CUB, FGVC Aircraft, Stanford Dogs, and Stanford Cars show that the proposed SGN outperforms the state-of-the-art WSOL methods.