IJCAI 2020 Papers

Text summarization is the research area aiming at creating a short and condensed version of the original document, which conveys the main idea of the document in a few words. This research topic has started to attract the attention of a large community of researchers, and it is nowadays counted as one of the most promising research areas. In general, text summarization algorithms aim at using a plain text document as input and then output a summary. However, in real-world applications, most of the data is not in a plain text format. Instead, there is much manifold information to be summarized, such as the summary for a web page based on a query in the search engine, extreme long document (e.g. academic paper), dialog history and so on. In this paper, we focus on the survey of these new summarization tasks and approaches in the real-world application.

As one of fundamental properties to characterize inconsistency measures for knowledge bases, the property of free formula independence captures well the intuition that free formulas are independent of the amount of inconsistency in a knowledge base for cases where inconsistency is characterized in terms of minimal inconsistent subsets. But it has been argued that not all the free formulas are independent of inconsistency in some other contexts of inconsistency characterization. In this paper, we propose a notion of Bi-free formula to describe formulas that are free from inconsistency in both syntactic characterization and paraconsistent models in the framework of Priest's minimally inconsistent LP. Then we propose the property of Bi-free formula independence, which is more suitable for characterizing the role of formulas free from inconsistency in measuring inconsistency from both syntactic and semantic perspectives.

We present a formal framework that allows individual and group of agents to reason about their trust toward other agents. In particular, we propose a branching time temporal logic BT which includes operators that express concepts such as everyone trust, distributed trust and propagated trust. We analyze the satisfiability and model checking problems of this logic using a reduction technique.

Formal query building is an important part of complex question answering over knowledge bases. It aims to build correct executable queries for questions. Recent methods try to rank candidate queries generated by a state-transition strategy. However, this candidate generation strategy ignores the structure of queries, resulting in a considerable number of noisy queries. In this paper, we propose a new formal query building approach that consists of two stages. In the first stage, we predict the query structure of the question and leverage the structure to constrain the generation of the candidate queries. We propose a novel graph generation framework to handle the structure prediction task and design an encoder-decoder model to predict the argument of the predetermined operation in each generative step. In the second stage, we follow the previous methods to rank the candidate queries. The experimental results show that our formal query building approach outperforms existing methods on complex questions while staying competitive on simple questions.

This work tackles the problem of checking strong equivalence of logic programs that may contain local auxiliary atoms, to be removed from their stable models and to be forbidden in any external context. We call this property projective strong equivalence (PSE). It has been recently proved that not any logic program containing auxiliary atoms can be reformulated, under PSE, as another logic program or formula without them -- this is known as strongly persistent forgetting. In this paper, we introduce a conservative extension of Equilibrium Logic and its monotonic basis, the logic of Here-and-There, in which we deal with a new connective we call fork. We provide a semantic characterisation of PSE for forks and use it to show that, in this extension, it is always possible to forget auxiliary atoms under strong persistence. We further define when the obtained fork is representable as a regular formula.

Early forecasting of bird migration patterns has important application for example in reducing avian biodiversity loss. An estimated 100 million to 1 billion birds are known to die yearly during migration due to fatal collisions with human made infrastructures such as buildings, high tension lines, wind turbines and aircrafts thus raising a huge concern for conservationists. Building models that can forecast accurate migration patterns is therefore important to enable the optimal management of these critical infrastructures with the sole aim of reducing biodiversity loss. While previous works have largely focused on the task of forecasting migration intensities and the onset of just one migration state, predicting several migration states at even finer granularity is more useful towards optimally managing the infrastructures that causes these deaths. In this work, we consider the task of forecasting migration patterns of the popular Turkey Vulture (Cathartes aura) collected with the aid of satellite telemetry for multiple years at a resolution of one hour. We use a deep Bidirectional-GRU recurrent neural network augmented with an auxiliary task where the state information of one layer is used to initialise the other. Empirical results on a variety of experiments with our approach show we can accurately forecast migration up to one week in advance performing better than a variety of baselines.

Audio synthesizers are pervasive in modern music production. These highly complex audio generation functions provide a unique diversity through their large sets of parameters. However, this feature also can make them extremely hard and obfuscated to use, especially for non-expert users with no formal knowledge on signal processing. We recently introduced a novel formalization of the problem of synthesizer control as learning an invertible mapping between an audio latent space, extracted from the audio signal, and a target parameter latent space, extracted from the synthesizer's presets, using normalizing flows. In addition to model a continuous representation allowing to ease the intuitive exploration of the synthesizer, it also provides a ground-breaking method for audio-based parameter inference, vocal control and macro-control learning. Here, we discuss the details of integrating these high-level features to develop new interaction schemes between a human user and the generating device: parameters inference from audio, high-level preset visualization and interpolation, that can be used both in off-time and real-time situations. Moreover, we also leverage LeapMotion devices to allow the control of hundreds of parameters simply by moving one hand across space to explore the low-dimensional latent space, allowing to both empower and facilitate the user's interaction with the synthesizer.

In this paper, we focus on a prediction-based novelty estimation strategy upon the deep reinforcement learning (DRL) framework, and present a flow-based intrinsic curiosity module (FICM) to exploit the prediction errors from optical flow estimation as exploration bonuses. We propose the concept of leveraging motion features captured between consecutive observations to evaluate the novelty of observations in an environment. FICM encourages a DRL agent to explore observations with unfamiliar motion features, and requires only two consecutive frames to obtain sufficient information when estimating the novelty. We evaluate our method and compare it with a number of existing methods on multiple benchmark environments, including Atari games, Super Mario Bros., and ViZDoom. We demonstrate that FICM is favorable to tasks or environments featuring moving objects, which allow FICM to utilize the motion features between consecutive observations. We further ablatively analyze the encoding efficiency of FICM, and discuss its applicable domains comprehensively. See here for our codes and demo videos.

Network Creation Games(NCGs) model the creation of decentralized communication networks like the Internet. In such games strategic agents corresponding to network nodes selfishly decide with whom to connect to optimize some objective function. Past research intensively analyzed models where the agents strive for a central position in the network. This models agents optimizing the network for low-latency applications like VoIP. However, with today's abundance of streaming services it is important to ensure that the created network can satisfy the increased bandwidth demand. To the best of our knowledge, this natural problem of the decentralized strategic creation of networks with sufficient bandwidth has not yet been studied. We introduce Flow-Based NCGs where the selfish agents focus on bandwidth instead of latency. In essence, budget-constrained agents create network links to maximize their minimum or average network flow value to all other network nodes. Equivalently, this can also be understood as agents who create links to increase their connectivity and thus also the robustness of the network. For this novel type of NCG we prove that pure Nash equilibria exist, we give a simple algorithm for computing optimal networks, we show that the Price of Stability is 1 and we prove an (almost) tight bound of 2 on the Price of Anarchy. Last but not least, we show that our models do not admit a potential function.

Given a set of individuals qualifying or disqualifying each other, group identification is the task of identifying a socially qualified subgroup of individuals. Social qualification depends on the specific rule used to aggregate individual qualifications. The bribery problem in this context asks how many agents need to change their qualifications in order to change the outcome. Complementing previous results showing polynomial-time solvability or NP-hardness of bribery for various social rules in the constructive (aiming at making specific individuals socially qualified) or destructive (aiming at making specific individuals socially disqualified) setting, we provide a comprehensive picture of the parameterized computational complexity landscape. Conceptually, we also consider a more fine-grained concept of bribery cost, where we ask how many single qualifications need to be changed, and a more general bribery goal that combines the constructive and destructive setting.

Modern day trading practice resembles a thought experiment, where investors imagine various possibilities of future stock market and invest accordingly. Generative adversarial network (GAN) is highly relevant to this trading practice in two ways. First, GAN generates synthetic data by a neural network that is technically indistinguishable from the reality, which guarantees the reasonableness of the experiment. Second, GAN generates multitudes of fake data, which implements half of the experiment. In this paper, we present a new architecture of GAN and adapt it to portfolio risk minimization problem by adding a regression network to GAN (implementing the second half of the experiment). The new architecture is termed GANr. Battling against two distinctive networks: discriminator and regressor, GANr's generator aims to simulate a stock market that is close to the reality while allow for all possible scenarios. The resulting portfolio resembles a robust portfolio with data-driven ambiguity. Our empirical studies show that GANr portfolio is more resilient to bleak financial scenarios than CLSGAN and LASSO portfolios.

Financial risk is an essential indicator of investment, which can help investors to understand the market and companies better. Among the many influencing factors of financial risk, researchers find the earnings conference call is the most significant one. Predicting financial volatility after the earnings conference call has been critical to beneficiaries, including investors and company managers. However, previous work mainly focuses on the feature extraction from the word-level or document-level.The vital structure of conferences, the alternate dialogue, is ignored. In this paper, we introduced our Multi-Round Q&A Attention Network, which brings into account the dialogue form in the first place. Based on the data of earnings call transcripts, we apply our model to extract features of each round of dialogue through a bidirectional attention mechanism and predict the volatility after the earnings conference call events. The results prove that our model significantly outperforms the previous state-of-the-art methods and other baselines in three different periods.

Small and Medium-sized Enterprises (SMEs) are playing a vital role in the modern economy. Recent years, financial risk analysis for SMEs attracts lots of attentions from financial institutions. However, the financial risk analysis for SMEs usually suffers data deficiency problem, especially for the mobile financial institutions which seldom collect credit-related data directly from SMEs. Fortunately, although credit-related information of SMEs is hard to be acquired sufficiently, the interactive relationships between SMEs, which may contain valuable information of financial risk, is usually available for the mobile financial institutions. Finding out credit-related relationship of SME from massive interactions helps comprehensively model the SMEs thus improve the performance of financial risk analysis. In this paper, tackling the data deficiency problem of financial risk analysis for SMEs, we propose an innovative financial risk analysis framework with graph-based supply chain mining. Specifically, to capture the credit-related topology structural and temporal variation information of SMEs, we design and employ a novel spatial-temporal aware graph neural network, to mine supply chain relationship on a SME graph, and then analysis the credit risk based on the mined supply chain graph. Experimental results on real-world financial datasets prove the effectiveness of our proposal for financial risk analysis for SMEs.

There is growing interest in the tasks of financial text mining. Over the past few years, the progress of Natural Language Processing (NLP) based on deep learning advanced rapidly. Significant progress has been made with deep learning showing promising results on financial text mining models. However, as NLP models require large amounts of labeled training data, applying deep learning to financial text mining is often unsuccessful due to the lack of labeled training data in financial fields. To address this issue, we present FinBERT (BERT for Financial Text Mining) that is a domain specific language model pre-trained on large-scale financial corpora. In FinBERT, different from BERT, we construct six pre-training tasks covering more knowledge, simultaneously trained on general corpora and financial domain corpora, which can enable FinBERT model better to capture language knowledge and semantic information. The results show that our FinBERT outperforms all current state-of-the-art models. Extensive experimental results demonstrate the effectiveness and robustness of FinBERT. The source code and pre-trained models of FinBERT are available online.

Standard disaster response involves using drones (or helicopters) for reconnaissance and using people on the ground to mitigate the damage. In this paper, we look at the problem of wildfires and propose an efficient resource allocation strategy to cope with both dynamically changing environment and uncertainty. In particular, we propose Firefly, a new resource allocation algorithm, that can provably achieve optimal or near optimal solutions with high probability by first efficiently allocating observation drones to collect information to reduce uncertainty, and then allocate the firefighting units to extinguish fire. For the former, Firefly uses a combination of maximum set coverage formulation and a novel utility estimation technique, and it uses a knapsack formulation to calculate the allocation for the latter. We also demonstrate empirically by using a real-world dataset that Firefly achieves up to 80-90% performance of the offline optimal solution, even with a small amount of drones, in most of the cases.

Few-shot learning aims to learn a model that can be readily adapted to new unseen classes (concepts) by accessing one or few examples. Despite the successful progress, most of the few-shot learning approaches, concentrating on either global or local characteristics of examples, still suffer from weak generalization abilities. Inspired by the inverted pyramid theory, to address this problem, we propose an inverted pyramid network (IPN) that intimates the human's coarse-to-fine cognition paradigm. The proposed IPN consists of two consecutive stages, namely global stage and local stage. At the global stage, a class-sensitive contextual memory network (CCMNet) is introduced to learn discriminative support-query relation embeddings and predict the query-to-class similarity based on the contextual memory. Then at the local stage, a fine-grained calibration is further appended to complement the coarse relation embeddings, targeting more precise query-to-class similarity evaluation. To the best of our knowledge, IPN is the first work that simultaneously integrates both global and local characteristics in few-shot learning, approximately imitating the human cognition mechanism. Our extensive experiments on multiple benchmark datasets demonstrate the superiority of IPN, compared to a number of state-of-the-art approaches.

Human motion prediction is a task where we anticipate future motion based on past observation. Previous approaches rely on the access to large datasets of skeleton data, and thus are difficult to be generalized to novel motion dynamics with limited training data. In our work, we propose a novel approach named Motion Prediction Network (MoPredNet) for few-short human motion prediction. MoPredNet can be adapted to predicting new motion dynamics using limited data, and it elegantly captures long-term dependency in motion dynamics. Specifically, MoPredNet dynamically selects the most informative poses in the streaming motion data as masked poses. In addition, MoPredNet improves its encoding capability of motion dynamics by adaptively learning spatio-temporal structure from the observed poses and masked poses. We also propose to adapt MoPredNet to novel motion dynamics based on accumulated motion experiences and limited novel motion dynamics data. Experimental results show that our method achieves better performance over state-of-the-art methods in motion prediction.

Credit card transaction fraud costs billions of dollars to card issuers every year. Besides, the credit card transaction dataset is very skewed, there are much fewer samples of frauds than legitimate transactions. Due to the data security and privacy, different banks are usually not allowed to share their transaction datasets. These problems make traditional model difficult to learn the patterns of frauds and also difficult to detect them. In this paper, we introduce a novel framework termed as federated meta-learning for fraud detection. Different from the traditional technologies trained with data centralized in the cloud, our model enables banks to learn fraud detection model with the training data distributed on their own local database. A shared whole model is constructed by aggregating locallycomputed updates of fraud detection model. Banks can collectively reap the benefits of shared model without sharing the dataset and protect the sensitive information of cardholders. To achieve the good performance of classification, we further formulate an improved triplet-like metric learning, and design a novel meta-learning-based classifier, which allows joint comparison with K negative samples in each mini-batch. Experimental results demonstrate that the proposed approach achieves significantly higher performance compared with the other state-of-the-art approaches.

Building compact convolutional neural networks (CNNs) with reliable performance is a critical but challenging task, especially when deploying them in real-world applications. As a common approach to reduce the size of CNNs, pruning methods delete part of the CNN filters according to some metrics such as l1-norm. However, previous methods hardly leverage the information variance in a single feature map and the similarity characteristics among feature maps. In this paper, we propose a novel filter pruning method, which incorporates two kinds of feature map selections: diversity-aware selection (DFS) and similarity-aware selection (SFS). DFS aims to discover features with low information diversity while SFS removes features that have high similarities with others. We conduct extensive empirical experiments with various CNN architectures on publicly available datasets. The experimental results demonstrate that our model obtains up to 91.6% parameter decrease and 83.7% FLOPs reduction with almost no accuracy loss.