Alibaba Group

About: Alibaba Group is a company organization based out in Hangzhou, China. It is known for research contribution in the topics: Computer science & Terminal (electronics). The organization has 6810 authors who have published 7389 publications receiving 55653 citations. The organization is also known as: Alibaba Group Holding Limited & Alibaba Group (Cayman Islands).

Unsupervised Deep Spectrum Sensing: A Variational Auto-Encoder Based Approach

Abstract: In cognitive radio (CR), the test statistics of most spectrum sensing algorithms are generated from the model-based features such as the signal energy and the eigenvalues from the sample covariance matrix (CM). Despite their low complexity, their detection performance depends very much on the accuracy of the presumed model. Also, these model-based statistics may not be able to exploit the full potential of the signal samples. To this end, the data-driven deep learning-based detectors have been proposed, with test statistics generated directly from signal samples in an automatic manner. However, existing deep learning-based detectors are all supervised learning-based and they usually require a massive amount of labeled training data to achieve decent detection performance. In practical CR scenarios, however, obtaining a large amount of labeled training data may be difficult. To address this issue, in this paper, we propose an unsupervised deep learning based spectrum sensing method named unsupervised deep spectrum sensing (UDSS). The UDSS algorithm requires no prior information such as the noise power or the signal's statistical CM. Moreover, the UDSS only requires a small amount of samples collected in absence of the primary user's (PU) signals ( $H_0$ labeled data). Simulation results show that the proposed UDSS algorithm is able to approach the performance of the benchmark deep supervised learning-based spectrum sensing algorithm and outperforms the model-based benchmark algorithms under both Gaussian noise and Laplace noise.

Ride Comfort Optimization via Speed Planning and Preview Semi-Active Suspension Control for Autonomous Vehicles on Uneven Roads

Abstract: By simultaneously utilizing preview and global road information, a comfort optimization strategy which combines vehicle speed planning and preview semi-active suspension control is designed for autonomous vehicles. Considering that the impact of vehicle speed at the suspension vibration source is always a barrier for preview suspension control, a processing method for the road data is novelly proposed. Then, to utilize the processed data and to handle the nonlinearity of semi-active actuators, a hybrid horizon-varying (HV) model predictive control (MPC) method is given. The method can adapt to speed variation and meanwhile take the most of the road data within a fixed preview length. Further, based on the global information and considering multiple road irregularities in a driving path, a speed planning problem is established in the spatial domain and a dynamic programming based solution is provided. The final speed trajectory can compromise the driving time, vertical vibration and longitudinal acceleration. Various simulation results have been employed to verify the superiority of the hybrid HV-MPC method and the significance of speed and suspension coordination for comfort improvement.

MemCap: Memorizing Style Knowledge for Image Captioning

Abstract: Generating stylized captions for images is a challenging task since it requires not only describing the content of the image accurately but also expressing the desired linguistic style appropriately. In this paper, we propose MemCap, a novel stylized image captioning method that explicitly encodes the knowledge about linguistic styles with memory mechanism. Rather than relying heavily on a language model to capture style factors in existing methods, our method resorts to memorizing stylized elements learned from training corpus. Particularly, we design a memory module that comprises a set of embedding vectors for encoding style-related phrases in training corpus. To acquire the style-related phrases, we develop a sentence decomposing algorithm that splits a stylized sentence into a style-related part that reflects the linguistic style and a content-related part that contains the visual content. When generating captions, our MemCap first extracts content-relevant style knowledge from the memory module via an attention mechanism and then incorporates the extracted knowledge into a language model. Extensive experiments on two stylized image captioning datasets (SentiCap and FlickrStyle10K) demonstrate the effectiveness of our method.

Joint Optimization of Tree-based Index and Deep Model for Recommender Systems

Abstract: Large-scale industrial recommender systems are usually confronted with computational problems due to the enormous corpus size. To retrieve and recommend the most relevant items to users under response time limits, resorting to an efficient index structure is an effective and practical solution. The previous work Tree-based Deep Model (TDM) \cite{zhu2018learning} greatly improves recommendation accuracy using tree index. By indexing items in a tree hierarchy and training a user-node preference prediction model satisfying a max-heap like property in the tree, TDM provides logarithmic computational complexity w.r.t. the corpus size, enabling the use of arbitrary advanced models in candidate retrieval and recommendation. In tree-based recommendation methods, the quality of both the tree index and the user-node preference prediction model determines the recommendation accuracy for the most part. We argue that the learning of tree index and preference model has interdependence. Our purpose, in this paper, is to develop a method to jointly learn the index structure and user preference prediction model. In our proposed joint optimization framework, the learning of index and user preference prediction model are carried out under a unified performance measure. Besides, we come up with a novel hierarchical user preference representation utilizing the tree index hierarchy. Experimental evaluations with two large-scale real-world datasets show that the proposed method improves recommendation accuracy significantly. Online A/B test results at a display advertising platform also demonstrate the effectiveness of the proposed method in production environments.

Understanding and bridging the gaps in current GNN performance optimizations

Abstract: Graph Neural Network (GNN) has recently drawn a rapid increase of interest in many domains for its effectiveness in learning over graphs. Maximizing its performance is essential for many tasks, but remains preliminarily understood. In this work, we provide an in-depth examination of the state-of-the-art GNN frameworks, revealing five major gaps in the current frameworks in optimizing GNN performance, especially in handling the special complexities of GNN over traditional graph or DNN operations. Based on the insights, we put together a set of optimizations to fill the gaps. These optimizations leverage the state-of-the-art GPU optimization techniques and tailor them to the special properties of GNN. Experimental results show that these optimizations achieve 1.37×--15.5× performance improvement over the state-of-the-art frameworks on various GNN models.