IJCAI 2020 论文列表

In peer selection agents must choose a subset of themselves for an award or a prize. As agents are self-interested, we want to design algorithms that are impartial, so that an individual agent cannot affect their own chance of being selected. This problem has broad application in resource allocation and mechanism design and has received substantial attention in the artificial intelligence literature. Here, we present a novel algorithm for impartial peer selection, PeerNomination, and provide a theoretical analysis of its accuracy. Our algorithm possesses various desirable features. In particular, it does not require an explicit partitioning of the agents, as previous algorithms in the literature. We show empirically that it achieves higher accuracy than the exiting algorithms over several metrics.

We derive conditions under which a peer-consistency mechanism can be used to elicit truthful data from non-trusted rational agents when an aggregate statistic of the collected data affects the amount of their incentives to lie. Furthermore, we discuss the relative saving that can be achieved by the mechanism, compared to the rational outcome, if no such mechanism was implemented. Our work is motivated by distributed platforms, where decentralized data oracles collect information about real-world events, based on the aggregate information provided by often self-interested participants. We compare our theoretical observations with numerical simulations on two public real datasets.

We present an attention-weighted loss in a photometric stereo neural network to improve 3D surface recovery accuracy in complex-structured areas, such as edges and crinkles, where existing learning-based methods often failed. Instead of using a uniform penalty for all pixels, our method employs the attention-weighted loss learned in a self-supervise manner for each pixel, avoiding blurry reconstruction result in such difficult regions. The network first estimates a surface normal map and an adaptive attention map, and then the latter is used to calculate a pixel-wise attention-weighted loss that focuses on complex regions. In these regions, the attention-weighted loss applies higher weights of the detail-preserving gradient loss to produce clear surface reconstructions. Experiments on real datasets show that our approach significantly outperforms traditional photometric stereo algorithms and state-of-the-art learning-based methods.

We demonstrate a pattern-based MIDI music generation system with a generation strategy based on Wasserstein autoencoders and a novel variant of pianoroll descriptions of patterns which employs separate channels for note velocities and note durations and can be fed into classic DCGAN-style convolutional architectures. We trained the system on two new datasets (in the acid-jazz and high-pop genres) composed by musicians in our team with music generation in mind. Our demonstration shows that moving smoothly in the latent space allows us to generate meaningful sequences of four-bars patterns.

Participatory budgeting systems allow city residents to jointly decide on projects they wish to fund using public money, by letting residents vote on such projects. While participatory budgeting is gaining popularity, existing aggregation methods do not take into account the natural possibility of project interactions, such as substitution and complementarity effects. Here we take a step towards fixing this issue: First, we augment the standard model of participatory budgeting by introducing a partition over the projects and model the type and extent of project interactions within each part using certain functions. We study the computational complexity of finding bundles that maximize voter utility, as defined with respect to such functions. Motivated by the desire to incorporate project interactions in real-world participatory budgeting systems, we identify certain cases that admit efficient aggregation in the presence of such project interactions.

Partial Multi-Label Learning (PML) aims to learn from the training data where each instance is associated with a set of candidate labels, among which only a part of them are relevant. Existing PML methods mainly focus on label disambiguation, while they lack the consideration of noise in the feature space. To tackle the problem, we propose a novel framework named partial multi-label learning via MUlti-SubspacE Representation (MUSER), where the redundant labels together with noisy features are jointly taken into consideration during the training process. Specifically, we first decompose the original label space into a latent label subspace and a label correlation matrix to reduce the negative effects of redundant labels, then we utilize the correlations among features to project the original noisy feature space to a feature subspace to resist the noisy feature information. Afterwards, we introduce a graph Laplacian regularization to constrain the label subspace to keep intrinsic structure among features and impose an orthogonality constraint on the correlations among features to guarantee discriminability of the feature subspace. Extensive experiments conducted on various datasets demonstrate the superiority of our proposed method.

Learning attacker behavior is an important research topic in security games as security agencies are often uncertain about attackers' decision making. Previous work has focused on developing various behavioral models of attackers based on historical attack data. However, a clever attacker can manipulate its attacks to fail such attack-driven learning, leading to ineffective defense strategies. We study attacker behavior deception with three main contributions. First, we propose a new model, named partial behavior deception model, in which there is a deceptive attacker (among multiple attackers) who controls a portion of attacks. Our model captures real-world security scenarios such as wildlife protection in which multiple poachers are present. Second, we introduce a new scalable algorithm, GAMBO, to compute an optimal deception strategy of the deceptive attacker. Our algorithm employs the projected gradient descent and uses the implicit function theorem for the computation of gradient. Third, we conduct a comprehensive set of experiments, showing a significant benefit for the attacker and loss for the defender due to attacker deception.

One-class novelty detection is to identify anomalous instances that do not conform to the expected normal instances. In this paper, the Generative Adversarial Networks (GANs) based on encoder-decoder-encoder pipeline are used for detection and achieve state-of-the-art performance. However, deep neural networks are too over-parameterized to deploy on resource-limited devices. Therefore, Progressive Knowledge Distillation with GANs (P-KDGAN) is proposed to learn compact and fast novelty detection networks. The P-KDGAN is a novel attempt to connect two standard GANs by the designed distillation loss for transferring knowledge from the teacher to the student. The progressive learning of knowledge distillation is a two-step approach that continuously improves the performance of the student GAN and achieves better performance than single step methods. In the first step, the student GAN learns the basic knowledge totally from the teacher via guiding of the pre-trained teacher GAN with fixed weights. In the second step, joint fine-training is adopted for the knowledgeable teacher and student GANs to further improve the performance and stability. The experimental results on CIFAR-10, MNIST, and FMNIST show that our method improves the performance of the student GAN by 2.44%, 1.77%, and 1.73% when compressing the computation at ratios of 24.45:1, 311.11:1, and 700:1, respectively.

The inherent heavy computation of deep neural networks prevents their widespread applications. A widely used method for accelerating model inference is quantization, by replacing the input operands of a network using fixed-point values. Then the majority of computation costs focus on the integer matrix multiplication accumulation. In fact, high-bit accumulator leads to partially wasted computation and low-bit one typically suffers from numerical overflow. To address this problem, we propose an overflow aware quantization method by designing trainable adaptive fixed-point representation, to optimize the number of bits for each input tensor while prohibiting numeric overflow during the computation. With the proposed method, we are able to fully utilize the computing power to minimize the quantization loss and obtain optimized inference performance. To verify the effectiveness of our method, we conduct image classification, object detection, and semantic segmentation tasks on ImageNet, Pascal VOC, and COCO datasets, respectively. Experimental results demonstrate that the proposed method can achieve comparable performance with state-of-the-art quantization methods while accelerating the inference process by about 2 times.

Answer Set Programming (ASP) is a well-known formalism for Knowledge Representation and Reasoning, successfully employed to solve many AI problems, also thanks to the availability of efficient implementations. Traditionally, ASP systems are based on the ground&solve approach, where the grounding transforms a general input program into its propositional counterpart, whose stable models are then computed by the solver using the CDCL algorithm. This approach suffers an intrinsic limitation: the grounding of one or few constraints may be unaffordable from a computational point of view; a problem known as grounding bottleneck. In this paper, we develop an innovative approach for evaluating ASP programs, where some of the constraints of the input program are not grounded but automatically translated into propagators of the CDCL algorithm that work on partial interpretations. We implemented the new approach on top of the solver WASP and carried out an experimental analysis on different benchmarks. Results show that our approach consistently outperforms state-of-the-art ASP systems by overcoming the grounding bottleneck.

Most Visual Question Answering (VQA) models suffer from the language prior problem, which is caused by inherent data biases. Specifically, VQA models tend to answer questions (e.g., what color is the banana?) based on the high-frequency answers (e.g., yellow) ignoring image contents. Existing approaches tackle this problem by creating delicate models or introducing additional visual annotations to reduce question dependency and strengthen image dependency. However, they are still subject to the language prior problem since the data biases have not been fundamentally addressed. In this paper, we introduce a self-supervised learning framework to solve this problem. Concretely, we first automatically generate labeled data to balance the biased data, and then propose a self-supervised auxiliary task to utilize the balanced data to assist the VQA model to overcome language priors. Our method can compensate for the data biases by generating balanced data without introducing external annotations. Experimental results show that our method achieves state-of-the-art performance, improving the overall accuracy from 49.50% to 57.59% on the most commonly used benchmark VQA-CP v2. In other words, we can increase the performance of annotation-based methods by 16% without using external annotations. Our code is available on GitHub.

In multivariate event data, the instantaneous rate of an event's occurrence may be sensitive to the temporal sequence in which other influencing events have occurred in the history. For example, an agent’s actions are typically driven by preceding actions taken by the agent as well as those of other relevant agents in some order. We introduce a novel statistical/causal model for capturing such an order-sensitive historical dependence, where an event’s arrival rate is determined by the order in which its underlying causal events have occurred in the recent past. We propose an algorithm to discover these causal events and learn the most influential orders using time-stamped event occurrence data. We show that the proposed model fits various event datasets involving single as well as multiple agents better than baseline models. We also illustrate potentially useful insights from our proposed model for an analyst during the discovery process through analysis on a real-world political event dataset.

Numerous tasks at the core of statistics, learning, and vision areas are specific cases of ill-posed inverse problems. Recently, learning-based (e.g., deep) iterative methods have been empirically shown to be useful for these problems. Nevertheless, integrating learnable structures into iterations is still a laborious process, which can only be guided by intuitions or empirical insights. Moreover, there is a lack of rigorous analysis of the convergence behaviors of these reimplemented iterations, and thus the significance of such methods is a little bit vague. We move beyond these limits and propose a theoretically guaranteed optimization learning paradigm, a generic and provable paradigm for nonconvex inverse problems, and develop a series of convergent deep models. Our theoretical analysis reveals that the proposed optimization learning paradigm allows us to generate globally convergent trajectories for learning-based iterative methods. Thanks to the superiority of our framework, we achieve state-of-the-art performance on different real applications.

This work investigates the problem of optimising a partial-order plan’s (POP) flexibility through the simultaneous transformation of its action ordering and variable binding constraints. While the former has been extensively studied through the notions of deordering and reordering, the latter has received much less attention. We show that a plan’s variable bindings are often related to resource usage and their reinstantiation can yield more flexible plans. To do so, we extend existing POP optimality criteria to support variable reinstantiation, and prove that checking if a plan can be optimised further is NP-complete. We also propose a MaxSAT-based technique for increasing plan flexibility and provide a thorough experimental evaluation that suggests that there are benefits in action reinstantiation.

Functional dependency can lead to discoveries of new mechanisms not possible via symmetric association. Most asymmetric methods for causal direction inference are not driven by the function-versus-independence question. A recent exact functional test (EFT) was designed to detect functionally dependent patterns model-free with an exact null distribution. However, the EFT lacked a theoretical justification, had not been compared with other asymmetric methods, and was practically slow. Here, we prove the functional optimality of the EFT statistic, demonstrate its advantage in functional inference accuracy over five other methods, and develop a branch-and-bound algorithm with dynamic and quadratic programming to run at orders of magnitude faster than its previous implementation. Our results make it practical to answer the exact functional dependency question arising from discovery-driven artificial intelligence applications. Software that implements EFT is freely available in the R package 'FunChisq' (≥2.5.0) at https://cran.r-project.org/package=FunChisq

This paper presents CORTA, a software agent that designs personalized rehabilitation programs for homeless youth suffering from opioid addiction. Many rehabilitation centers treat opioid addiction in homeless youth by prescribing rehabilitation programs that are tailored to the underlying causes of addiction. To date, rehabilitation centers have relied on ad-hoc assessments and unprincipled heuristics to deliver rehabilitation programs to homeless youth suffering from opioid addiction, which greatly undermines the effectiveness of the delivered programs. CORTA addresses these challenges via three novel contributions. First, CORTA utilizes a first-of-its-kind real-world dataset collected from ~1400 homeless youth to build causal inference models which predict the likelihood of opioid addiction among these youth. Second, utilizing counterfactual predictions generated by our causal inference models, CORTA solves novel optimization formulations to assign appropriate rehabilitation programs to the correct set of homeless youth in order to minimize the expected number of homeless youth suffering from opioid addiction. Third, we provide a rigorous experimental analysis of CORTA along different dimensions, e.g., importance of causal modeling, importance of optimization, and impact of incorporating fairness considerations, etc. Our simulation results show that CORTA outperforms baselines by ~110% in minimizing the number of homeless youth suffering from opioid addiction.

Region search is an important problem in location-based services due to its wide applications. In this paper, we study the problem of optimal region search with submodular maximization (ORS-SM). This problem considers a region as a connected subgraph. We compute an objective value over the locations in the region using a submodular function and a budget value by summing up the costs of edges in the region, and aim to search the region with the largest objective score under a budget value constraint. ORS-SM supports many applications such as the most diversified region search. We prove that the problem is NP-hard and develop two approximation algorithms with guaranteed error bounds. We conduct experiments on two applications using three real-world datasets. The results demonstrate that our algorithms can achieve high-quality solutions and are faster than a state-of-the-art method by orders of magnitude.

With the recent advances in both machine learning and embedded systems research, the demand to deploy computational models for real-time execution on edge devices has increased substantially. Without deploying computational models on edge devices, the frequent transmission of sensor data to the cloud results in rapid battery draining due to the energy consumption of wireless data transmission. This rapid power dissipation leads to a considerable reduction in the battery lifetime of the system, therefore jeopardizing the real-world utility of smart devices. It is well-established that for difficult machine learning tasks, models with higher performance often require more computation power and thus are not power-efficient choices for deployment on edge devices. However, the trade-offs between performance and power consumption are not well studied. While numerous methods (e.g., model compression) have been developed to obtain an optimal model, these methods focus on improving the efficiency of a "single" model. In an entirely new direction, we introduce an effective method to find a combination of "multiple" models that are optimal in terms of power-efficiency and performance by solving an optimization problem in which both performance and power consumption are taken into account. Experimental results demonstrate that on the ImageNet dataset, we can achieve a 20% energy reduction with only 0.3% accuracy drop compared to Squeeze-and-Excitation Networks. Compared to a pruned convolutional neural network for human activity recognition, while consuming 1.7% less energy, our proposed policy achieves 1.3% higher accuracy.

We consider the problem of planning with arithmetic theories, and focus on generating optimal plans for numeric domains with constant and state-dependent action costs. Solving these problems efficiently requires a seamless integration between propositional and numeric reasoning. We propose a novel approach that leverages Optimization Modulo Theories (OMT) solvers to implement a domain-independent optimal theory-planner. We present a new encoding for optimal planning in this setting and we evaluate our approach using well-known, as well as new, numeric benchmarks.