With the outbreak of COVID-19, it is urgent and necessary to design a system that can access to information from COVID-19 related documents. Current methods fail to do so since the knowledge about COVID-19, an emerging disease, keeps changing and growing. In this study, we design a dynamic document-based question answering system, namely Web Understanding and Learning with AI (WULAI-QA). WULAI-QA employs feature engineering and online learning to adapt to the non-stationary environment and maintains good and steady performance. We evaluate WULAI-QA's performance on a public question answering (https://www.datafountain.cn/competitions/424) and rank first. We demonstrate that WULAI-QA can learn from user feedback and is easy to use. We believe that WULAI-QA will definitely help people understand COVID-19 and play an important role to fight against the pandemic.

Since late December 2019, it has been reported an outbreak of atypical pneumonia, now known as COVID-19 caused by the novel coronavirus. Cases have spread to more than 200 countries and regions internationally. World Health Organization (WHO) officially declares the coronavirus outbreak a pandemic and the public health emergency has caused world-wide impact to daily lives: people are advised to keep social distance, in-person events have been moved online, and some function facilitates have been locked-down. Alternatively, the Web becomes an active venue for people to share information. With respect to the on-going topic, people continuously post questions online and seek for answers. Yet, sharing global information conveyed in different languages is challenging because the language barrier is intrinsically unfriendly to monolingual speakers. In this paper, we propose a multilingual COVID-QA model to answer people’s questions in their own languages while the model is able to absorb knowledge from other languages. Another challenge is that in most cases, the information to share does not have parallel data in multiple languages. To this end, we propose a novel framework which incorporates (unsupervised) translation alignment to learn as pseudo-parallel data. Then we train multilingual question-answering mapping and generation. We demonstrate the effectiveness of our proposed approach compared against a series of competitive baselines. In this way, we make it easier to share global information across the language barriers, and hopefully we contribute to the battle against COVID-19.

Dating back from late December 2019, the Chinese city of Wuhan has reported an outbreak of atypical pneumonia, now known as lung inflammation caused by novel coronavirus (COVID-19). Cases have spread to other cities in China and more than 180 countries and regions internationally. World Health Organization (WHO) officially declares the coronavirus outbreak a pandemic and the public health emergency is perhaps one of the top concerns in the year of 2020 for governments all over the world. Till today, the coronavirus outbreak is still raging and has no sign of being under control in many countries. In this paper, we aim at drawing lessons from the COVID-19 outbreak process in China and using the experiences to help the interventions against the coronavirus wherever in need. To this end, we have built a system predicting hazard areas on the basis of confirmed infection cases with location information. The purpose is to warn people to avoid of such hot zones and reduce risks of disease transmission through droplets or contacts. We analyze the data from the daily official information release which are publicly accessible. Based on standard classification frameworks with reinforcements incrementally learned day after day, we manage to conduct thorough feature engineering from empirical studies, including geographical, demographic, temporal, statistical, and epidemiological features. Compared with heuristics baselines, our method has achieved promising overall performance in terms of precision, recall, accuracy, F1 score, and AUC. We expect that our efforts could be of help in the battle against the virus, the common opponent of human kind.

Despite great strides made in facial verification, it remains challenging to match facial images across different modalities. This is mainly due to the cross-modal gap induced by feature heterogeneity. Much prior work had focused on bridging the feature gap, resulting in near-perfect matching accuracies for viewed sketches. Nonetheless, studies on matching unviewed (forensic) sketches and caricatures, a much harder problem due to the additional cross-modal gap introduced by visual abstraction, had only just commenced in recent years. In this paper, we focus on matching facial caricatures with photos by directly addressing the visual abstraction problem. We show that by synergizing a taxonomy of fine-grained visual attributes with part-aware low-level feature extraction, the visual abstraction gap can be effectively traversed, resulting in improved overall cross-modal matching accuracy. More specifically, (i) we propose a simple yet effective geometry-based attribute classifier to detect fine-grained attributes at part-level, and (ii) we demonstrate how meaningful facial regions can be reliably detected to enable localized feature extraction and attribute detection, and (iii) we show a common embedding can be learned using Canonical Correlation Analysis (CCA) that combines part-based low-level features and fine-grained visual attributes. We demonstrate the superiority of the proposed cross-modal strategy by evaluating on two recent photo-caricature datasets.

We study the problem of fine-grained sketch-based image retrieval. By performing instance-level (rather than category-level) retrieval, it embodies a timely and practical application, particularly with the ubiquitous availability of touchscreens. Three factors contribute to the challenging nature of the problem: (i) free-hand sketches are inherently abstract and iconic, making visual comparisons with photos more difficult, (ii) sketches and photos are in two different visual domains, i.e. black and white lines vs. color pixels, and (iii) fine-grained distinctions are especially challenging when executed across domain and abstraction-level. To address this, we propose to detect visual attributes at part-level, in order to build a new representation that not only captures fine-grained characteristics but also traverses across visual domains. More specifically, (i) we propose a dataset with 304 photos and 912 sketches, where each sketch and photo is annotated with its semantic parts and associated part-level attributes, and with the help of this dataset, we investigate (ii) how strongly-supervised deformable part-based models can be learned that subsequently enable automatic detection of part-level attributes, and (iii) a novel matching framework that synergistically integrates low-level features, mid-level geometric structure and high-level semantic attributes to boost retrieval performance. Extensive experiments conducted on our new dataset demonstrate value of the proposed method.

This paper considers algorithms for optimal transmission over wireless multiuser channels where the transmitter has access to imperfect channel state information (CSI). We focus on downlink orthogonal frequency division multiple access and multiuser uplink random access. In both cases, frequency assignment, transmitted power, and coding mode are adapted to imperfect CSI in order to maximize expected transmission rate subject to average power constraints. Determination of optimal solutions is a non-convex stochastic optimization problem with infinitely many variables. Exploiting its property of null duality gap, we show that optimal solutions are determined by optimal dual variables. This affords considerable simplification because the dual optimization problem is convex and finite-dimensional. Iterative algorithms that find the optimal operating point based on imperfect CSI without having access to the channels' probability distributions are further developed.

We propose a method to improve traditional character-based PPM text compression algorithm [1] for natural languages. Consider a text file as a sequence of alternating words and non-words, the basic idea of our algorithm is to encode nonwords and prefixes of words using character-based context models and encode suffixes of words using dictionary models. By using dictionary models, the algorithm can encode multiple characters as a whole, and thus enhance the compression efficiency. The advantages of the proposed algorithm are: 1) it does not require any text preprocessing, 2) it does not need any explicit codeword to identify switch between context and dictionary models, 3) it can be applied to any character-based PPM algorithms without incurring much additional computational cost.

This paper considers a wireless channel when the probability distribution of fading is unknown and the transmitter has access to imperfect channel state information (CSI). Transmitted power and coding mode are adapted to the available imperfect CSI through the use of a backoff function in order to maximize the expected transmission rate subject to an average power constraint. Determination of the optimal power allocation and backoff function is a non-convex stochastic optimization problem with infinitely many variables. Despite its non-convexity, the duality gap of this problem has been shown to be null. Exploiting this property, we show that the optimal power allocation and channel backoff functions are uniquely determined by the optimal dual variable. This affords considerable simplification because the dual optimization problem is convex and one-dimensional -- whereas the original primal problem is non-convex and infinite-dimensional. Iterative algorithms that find the optimal power and backoff function based on imperfect CSI without having access to the channel probability distribution are further developed. Numerical results are provided to corroborate theoretical findings.

We consider distributed algorithms to optimize random access multihop wireless networks in the presence of fading. Since the associated optimization problem is neither convex nor amenable to distributed implementation, we introduce a problem approximation. This approximation is still not convex, but it has zero duality gap and can be solved and decomposed into local subproblems in the dual domain. The solution method is through a stochastic subgradient descent algorithm that operates without knowledge of the fading's distribution and leads to an architecture composed of layers and layer interfaces. With limited amount of message passing among terminals and small computational cost, the proposed algorithm converges almost surely in an ergodic sense.

We investigate the performance of a discrete-time all-analog-processing joint source-channel coding system for the transmission of memoryless sources over average power constrained AWGN channels. First, N:1 bandwidth compression systems are analyzed and optimized. At the encoder, N samples of an i.i.d. source are directly mapped into one channel symbol using a non-linear curve. Different from previous work in the literature, we introduce an additional degree of freedom at the encoder, MMSE decoding instead of ML decoding is considered, and we focus on both high and low channel signal-to-noise ratio (CSNR) regions. By using MMSE decoding, the proposed system presents a performance very close to the theoretical limits, even at low CSNR, as long as the system parameters are properly optimized. Then, N:K bandwidth compression systems are constructed by parallel combination of an M:1 system and a 1:1 uncoded system, and the optimal power allocation between the two constituent systems is derived in order to maximize the overall output signal-to-distortion ratio (SDR). Finally, 1:2 bandwidth expansion systems using mapping functions similar to those used in 2:1 system are investigated. Different from digital systems, the proposed scheme does not require long block lengths to achieve good performance, and shows graceful degradation when the CSNR is lower than the one used for the design.

We develop adaptive scheduling and power control algorithms for random access in a multiple access channel where terminals acquire instantaneous channel state information but do not know the probability distribution of the channel. In each time slot, terminals measure the channel to the common access point. Based on the observed channel value, they determine whether to transmit or not and, if they decide to do so, adjust their transmitted power. We remark that there is no coordination between terminals and that adaptation is to the local channel value only. It is shown that the proposed algorithm almost surely maximizes a proportional fair utility while adhering to instantaneous and average power constraints. Important properties of the algorithm are low computational complexity and the ability to handle non-convex rate functions. Numerical results on a randomly generated network with heterogeneous users corroborate theoretical results.

We investigate the performance of a discrete-time all-analog-processing joint source channel coding system for the transmission of i.i.d. Gaussian and Laplacian sources over AWGN channels. In the encoder, two samples of an i.i.d. source are mapped into a channel symbol using a space-filling curve. Different from previous work in the literature, MMSE decoding instead of ML decoding is considered, and we focus on both high and low channel SNR regions. The main contribution of this paper is to show that the proposed system presents a performance very close to the theoretical limits, even at low SNR, as long as the curve parameters are properly optimized.

An enhanced shot change detection algorithm for soccer video analysis is proposed in this paper. Features extracted from reliable motion vectors (MVs) are used to enhance the accuracy and efficiency of color-based method. In order to eliminate unreliable MVs, a MV filtration scheme is designed based on the principle of block-based motion search. Then two features, named as proportion of reliable MVs and centrality of reliable MVs respectively, are extracted from the filtration results in each frame. The experiments on 180 minutes' soccer video demonstrate that our algorithm achieved a considerable improvement over the widely-adopted color-based method.