AAAI 2022 论文列表

Providing model explanations has gained significant popularity recently. In contrast with the traditional feature-level model explanations, concept-based explanations can provide explanations in the form of high-level human concepts. However, existing concept-based explanation methods implicitly follow a two-step procedure that involves human intervention. Specifically, they first need the human to be involved to define (or extract) the high-level concepts, and then manually compute the importance scores of these identified concepts in a post-hoc way. This laborious process requires significant human effort and resource expenditure due to manual work, which hinders their large-scale deployability. In practice, it is challenging to automatically generate the concept-based explanations without human intervention due to the subjectivity of defining the units of concept-based interpretability. In addition, due to its data-driven nature, the interpretability itself is also potentially susceptible to malicious manipulations. Hence, our goal in this paper is to free human from this tedious process, while ensuring that the generated explanations are provably robust to adversarial perturbations. We propose a novel concept-based interpretation method, which can not only automatically provide the prototype-based concept explanations but also provide certified robustness guarantees for the generated prototype-based explanations. We also conduct extensive experiments on real-world datasets to verify the desirable properties of the proposed method.

We present a multi-objective meta-search procedure that constructs candidate algorithms for state-space search puzzles like Rubik's cube. The candidate algorithms take the form of macro databases, i.e., rule tables that specify sequences of actions to perform in different states. Rules are repeatedly applied until the puzzle is solved. The objectives favor candidates that are god-like (solving the puzzle in fewer steps) and folk-like (having fewer rules in the macro database). We build each candidate with a non-deterministic rule table construction, and then optimize over the non-deterministic choice points to find candidates near the Pareto-optimal trades-offs between godliness and folksiness. We prove that the rule table construction is correct: it always terminates and solves every state at termination. This is verified empirically on the full 2x2x2 "pocket" cube, where correct (but unoptimized) constructions take under one hour and the total number of rules is less than 10% the number of possible states. We also empirically assess the multi-objective optimization on restricted variants of the cube with up to 29K possible states, showing relative improvements in the objectives between 14-20%. Avenues for scaling up the method in future work are discussed.

We address the problem of facial motion retargeting that aims to transfer facial motion from a 2D face image to 3D characters. Existing methods often formulate this problem as a 3D face reconstruction problem, which estimates the face attributes such as face identity and expression from face images. However, due to the lack of ground-truth labels for both identity and expression, most 3D-face reconstruction-based methods fail to capture the facial identity and expression accurately. As a result, these methods may not achieve promising performance. To address this, we propose an identity-consistent constraint to learn accurate identities by encouraging consistent identity prediction across multiple frames. Based on a more accurate identity, we are able to obtain a more accurate facial expression. Moreover, we further propose an expression-exclusive constraint to improve performance by avoiding the co-occurrence of contradictory expression units (e.g., ``brow lower'' vs. ``brow raise''). Extensive experiments on facial motion retargeting and 3D face reconstruction tasks demonstrate the superiority of the proposed method over existing methods. Our code and supplementary materials are available at https://github.com/deepmo24/CPEM.

There has been significant recent interest in quantum neural networks (QNNs), along with their applications in diverse domains. Current solutions for QNNs pose significant challenges concerning their scalability, ensuring that the postulates of quantum mechanics are satisfied and that the networks are physically realizable. The exponential state space of QNNs poses challenges for the scalability of training procedures. The no-cloning principle prohibits making multiple copies of training samples, and the measurement postulates lead to non-deterministic loss functions. Consequently, the physical realizability and efficiency of existing approaches that rely on repeated measurement of several copies of each sample for training QNNs are unclear. This paper presents a new model for QNNs that relies on band-limited Fourier expansions of transfer functions of quantum perceptrons (QPs) to design scalable training procedures. This training procedure is augmented with a randomized quantum stochastic gradient descent technique that eliminates the need for sample replication. We show that this training procedure converges to the true minima in expectation, even in the presence of non-determinism due to quantum measurement. Our solution has a number of important benefits: (i) using QPs with concentrated Fourier power spectrum, we show that the training procedure for QNNs can be made scalable; (ii) it eliminates the need for resampling, thus staying consistent with the no-cloning rule; and (iii) enhanced data efficiency for the overall training process since each data sample is processed once per epoch. We present a detailed theoretical foundation for our models and methods' scalability, accuracy, and data efficiency. We also validate the utility of our approach through a series of numerical experiments.

In the context of skeleton-based action recognition, graph convolutional networks (GCNs) have been rapidly developed, whereas convolutional neural networks (CNNs) have received less attention. One reason is that CNNs are considered poor in modeling the irregular skeleton topology. To alleviate this limitation, we propose a pure CNN architecture named Topology-aware CNN (Ta-CNN) in this paper. In particular, we develop a novel cross-channel feature augmentation module, which is a combo of map-attend-group-map operations. By applying the module to the coordinate level and the joint level subsequently, the topology feature is effectively enhanced. Notably, we theoretically prove that graph convolution is a special case of normal convolution when the joint dimension is treated as channels. This confirms that the topology modeling power of GCNs can also be implemented by using a CNN. Moreover, we creatively design a SkeletonMix strategy which mixes two persons in a unique manner and further boosts the performance. Extensive experiments are conducted on four widely used datasets, i.e. N-UCLA, SBU, NTU RGB+D and NTU RGB+D 120 to verify the effectiveness of Ta-CNN. We surpass existing CNN-based methods significantly. Compared with leading GCN-based methods, we achieve comparable performance with much less complexity in terms of the required GFLOPs and parameters.

Deep clustering (DC) leverages the representation power of deep architectures to learn embedding spaces that are optimal for cluster analysis. This approach filters out low-level information irrelevant for clustering and has proven remarkably successful for high dimensional data spaces. Some DC methods employ Generative Adversarial Networks (GANs), motivated by the powerful latent representations these models are able to learn implicitly. In this work, we propose HC-MGAN, a new technique based on GANs with multiple generators (MGANs), which have not been explored for clustering. Our method is inspired by the observation that each generator of a MGAN tends to generate data that correlates with a sub-region of the real data distribution. We use this clustered generation to train a classifier for inferring from which generator a given image came from, thus providing a semantically meaningful clustering for the real distribution. Additionally, we design our method so that it is performed in a top-down hierarchical clustering tree, thus proposing the first hierarchical DC method, to the best of our knowledge. We conduct several experiments to evaluate the proposed method against recent DC methods, obtaining competitive results. Last, we perform an exploratory analysis of the hierarchical clustering tree that highlights how accurately it organizes the data in a hierarchy of semantically coherent patterns.

We present a novel semidefinite programming (SDP) relaxation that enables tight and efficient verification of neural networks. The tightness is achieved by combining SDP relaxations with valid linear cuts, constructed by using the reformulation-linearisation technique (RLT). The computational efficiency results from a layerwise SDP formulation and an iterative algorithm for incrementally adding RLT-generated linear cuts to the verification formulation. The layer RLT-SDP relaxation here presented is shown to produce the tightest SDP relaxation for ReLU neural networks available in the literature. We report experimental results based on MNIST neural networks showing that the method outperforms the state-of-the-art methods while maintaining acceptable computational overheads. For networks of approximately 10k nodes (1k, respectively), the proposed method achieved an improvement in the ratio of certified robustness cases from 0% to 82% (from 35% to 70%, respectively).

Using natural-language feedback to guide image generation and manipulation can greatly lower the required efforts and skills. This topic has received increased attention in recent years through refinement of Generative Adversarial Networks (GANs); however, most existing works are limited to single-round interaction, which is not reflective of real world interactive image editing workflows. Furthermore, previous works dealing with multi-round scenarios are limited to predefined feedback sequences, which is also impractical. In this paper, we propose a novel framework for Text-based Interactive image generation and manipulation (TiGAN) that responds to users' natural-language feedback. TiGAN utilizes the powerful pre-trained CLIP model to understand users' natural-language feedback and exploits contrastive learning for a better text-to-image mapping. To maintain the image consistency during interactions, TiGAN generates intermediate feature vectors aligned with the feedback and selectively feeds these vectors to our proposed generative model. Empirical results on several datasets show that TiGAN improves both interaction efficiency and image quality while better avoids undesirable image manipulation during interactions.

Following up on purely theoretical work, we contribute further theoretical insights into adapting stable two-sided matchings to change. Moreover, we perform extensive empirical studies hinting at numerous practically useful properties. Our theoretical extensions include the study of new problems (that is, incremental variants of Almost Stable Marriage and Hospital Residents), focusing on their (parameterized) computational complexity and the equivalence of various change types (thus simplifying algorithmic and complexity-theoretic studies for various natural change scenarios). Our experimental findings reveal, for instance, that allowing the new matching to be blocked by a few pairs significantly decreases the difference between the old and the new matching.

When designing algorithms for finite-time-horizon episodic reinforcement learning problems, a common approach is to introduce a fictitious discount factor and use stationary policies for approximations. Empirically, it has been shown that the fictitious discount factor helps reduce variance, and stationary policies serve to save the per-iteration computational cost. Theoretically, however, there is no existing work on convergence analysis for algorithms with this fictitious discount recipe. This paper takes the first step towards analyzing these algorithms. It focuses on two vanilla policy gradient (VPG) variants: the first being a widely used variant with discounted advantage estimations (DAE), the second with an additional fictitious discount factor in the score functions of the policy gradient estimators. Non-asymptotic convergence guarantees are established for both algorithms, and the additional discount factor is shown to reduce the bias introduced in DAE and thus improve the algorithm convergence asymptotically. A key ingredient of our analysis is to connect three settings of Markov decision processes (MDPs): the finite-time-horizon, the average reward and the discounted settings. To our best knowledge, this is the first theoretical guarantee on fictitious discount algorithms for the episodic reinforcement learning of finite-time-horizon MDPs, which also leads to the (first) global convergence of policy gradient methods for finite-time-horizon episodic reinforcement learning.

We introduce the triangle-densest-K-subgraph problem (TDKS) for undirected graphs: given a size parameter K, compute a subset of K vertices that maximizes the number of induced triangles. The problem corresponds to the simplest generalization of the edge based densest-K-subgraph problem (DKS) to the case of higher-order network motifs. We prove that TDKS is NP-hard and is not amenable to efficient approximation, in the worst-case. By judiciously exploiting the structure of the problem, we propose a relaxation algorithm for the purpose of obtaining high-quality, sub-optimal solutions. Our approach utilizes the fact that the cost function of TDKS is submodular to construct a convex relaxation for the problem based on the Lovász extension for submodular functions. We demonstrate that our approaches attain state-of-the-art performance on real-world graphs and can offer substantially improved exploration of the optimal density-size curve compared to sophisticated approximation baselines for DKS. We use document summarization to showcase why TDKS is a useful generalization of DKS.

We study the second-price auction in which bidders have asymmetric information regarding the item’s value. Each bidder’s value for the item depends on a private component and a public component. While each bidder observes their own private component, they hold different and asymmetric information about the public component. We characterize the equilibrium of this auction game and study how the asymmetric bidder information affects their equilibrium bidding strategies. We also discover multiple surprisingly counter-intuitive equilibrium phenomena. For instance, a bidder may be better off if she is less informed regarding the public component. Conversely, a bidder may sometimes be worse off if she obtains more accurate estimation about the auctioned item. Our results suggest that efforts devoted by bidders to improve their value estimations, as widely seen in today’s online advertising auctions, may not always be to their benefit.

The Cumulative constraint greatly contributes to the success of constraint programming at solving scheduling problems. The SoftCumulative, a version of the Cumulative where overloading the resource incurs a penalty is, however, less studied. We introduce a checker and a filtering algorithm for the SoftCumulative, which are inspired by the powerful energetic reasoning rule for the Cumulative. Both algorithms can be used with classic linear penalty function, but also with a quadratic penalty function, where the penalty of overloading the resource increases quadratically with the amount of the overload. We show that these algorithms are more general than existing algorithms and vastly outperform a decomposition of the SoftCumulative in practice.

When aggregating logically interconnected judgements from n agents, the result might be logically inconsistent. This phenomenon is known as the doctrinal paradox, which plays a central role in the field of judgement aggregation. Previous work has mostly focused on the worst-case analysis of the doctrinal paradox, leading to many impossibility results. Little is known about its likelihood of occurrence in practical settings, except for the study under certain distributions by List in 2005. In this paper, we characterize the likelihood of the doctrinal paradox under a general and realistic model called semi-random social choice framework (proposed by Xia in 2020). In the framework, agents' ground truth judgements can be arbitrarily correlated, while the noises are independent. Our main theorem states that under mild conditions, the semi-random likelihood of the doctrinal paradox is either 0, exp(-Θ(n)), Θ(n^~(-0.5)) or Θ(1). This not only answers open questions by List in 2005, but also draws clear lines between situations with frequent paradoxes and with vanishing paradoxes.

The classic secretary problem concerns the problem of an employer facing a random sequence of candidates and making online hiring decisions to try to hire the best candidate. In this paper, we study a game-theoretic generalization of the secretary problem where a set of employers compete with each other to hire the best candidate. Different from previous secretary market models, our model assumes that the sequence of candidates arriving at each employer is uniformly random but independent from other sequences. We consider two versions of this secretary game where employers can have adaptive or non-adaptive strategies, and provide characterizations of the best response and Nash equilibrium of each game.

Hyperparameter optimization is a ubiquitous challenge in machine learning, and the performance of a trained model depends crucially upon their effective selection. While a rich set of tools exist for this purpose, there are currently no practical hyperparameter selection methods under the constraint of differential privacy (DP). We study honest hyperparameter selection for differentially private machine learning, in which the process of hyperparameter tuning is accounted for in the overall privacy budget. To this end, we i) show that standard composition tools outperform more advanced techniques in many settings, ii) empirically and theoretically demonstrate an intrinsic connection between the learning rate and clipping norm hyperparameters, iii) show that adaptive optimizers like DPAdam enjoy a significant advantage in the process of honest hyperparameter tuning, and iv) draw upon novel limiting behaviour of Adam in the DP setting to design a new and more efficient optimizer.

In logic-based knowledge representation, query answering has essentially replaced mere satisfiability checking as the inferencing problem of primary interest. For knowledge bases in the basic description logic ALC, the computational complexity of conjunctive query (CQ) answering is well known to be EXPTIME-complete and hence not harder than satisfiability. This does not change when the logic is extended by certain features (such as counting or role hierarchies), whereas adding others (inverses, nominals or transitivity together with role-hierarchies) turns CQ answering exponentially harder. We contribute to this line of results by showing the surprising fact that even extending ALC by just the Self operator – which proved innocuous in many other contexts – increases the complexity of CQ entailment to 2EXPTIME. As common for this type of problem, our proof establishes a reduction from alternating Turing machines running in exponential space, but several novel ideas and encoding tricks are required to make the approach work in that specific, restricted setting.

In multiwinner approval voting, the goal is to select k-member committees based on voters' approval ballots. A well-studied concept of proportionality in this context is the justified representation (JR) axiom, which demands that no large cohesive group of voters remains unrepresented. However, the JR axiom may conflict with other desiderata, such as coverage (maximizing the number of voters who approve at least one committee member) or social welfare (maximizing the number of approvals obtained by committee members). In this work, we investigate the impact of imposing the JR axiom (as well as the more demanding EJR axiom) on social welfare and coverage. Our approach is threefold: we derive worst-case bounds on the loss of welfare/coverage that is caused by imposing JR, study the computational complexity of finding 'good' committees that provide JR (obtaining a hardness result, an approximation algorithm, and an exact algorithm for one-dimensional preferences), and examine this setting empirically on several synthetic datasets.

Formulating real-world optimization problems often begins with making predictions from historical data (e.g., an optimizer that aims to recommend fast routes relies upon travel-time predictions). Typically, learning the prediction model used to generate the optimization problem and solving that problem are performed in two separate stages. Recent work has showed how such prediction models can be learned end-to-end by differentiating through the optimization task. Such methods often yield empirical improvements, which are typically attributed to end-to-end making better error tradeoffs than the standard loss function used in a two-stage solution. We refine this explanation and more precisely characterize when end-to-end can improve performance. When prediction targets are stochastic, a two-stage solution must make an a priori choice about which statistics of the target distribution to model---we consider expectations over prediction targets---while an end-to-end solution can make this choice adaptively. We show that the performance gap between a two-stage and end-to-end approach is closely related to the emph{price of correlation} concept in stochastic optimization and show the implications of some existing POC results for the predict-then-optimize problem. We then consider a novel and particularly practical setting, where multiple prediction targets are combined to obtain each of the objective function’s coefficients. We give explicit constructions where (1) two-stage performs unboundedly worse than end-to-end; and (2) two-stage is optimal. We use simulations to experimentally quantify performance gaps and identify a wide range of real-world applications from the literature whose objective functions rely on multiple prediction targets, suggesting that end-to-end learning could yield significant improvements.