AAAI 2023 论文列表

Automatic International Classification of Diseases (ICD) coding aims to assign multiple ICD codes to a medical note with an average of 3,000+ tokens. This task is challenging due to the high-dimensional space of multi-label assignment (155,000+ ICD code candidates) and the long-tail challenge - Many ICD codes are infrequently assigned yet infrequent ICD codes are important clinically. This study addresses the long-tail challenge by transforming this multi-label classification task into an autoregressive generation task. Specifically, we first introduce a novel pretraining objective to generate free text diagnosis and procedure descriptions using the SOAP structure, the medical logic physicians use for note documentation. Second, instead of directly predicting the high dimensional space of ICD codes, our model generates the lower dimension of text descriptions, which then infer ICD codes. Third, we designed a novel prompt template for multi-label classification. We evaluate our Generation with Prompt (GP) model with the benchmark of all code assignment (MIMIC-III-full) and few shot ICD code assignment evaluation benchmark (MIMIC-III-few). Experiments on MIMIC-III-few show that our model performs with a marco F1 30.2, which substantially outperforms the previous MIMIC-III-full SOTA model (marco F1 4.3) and the model specifically designed for few/zero shot setting (marco F1 18.7). Finally, we design a novel ensemble learner, a cross attention reranker with prompts, to integrate previous SOTA and our best few-shot coding predictions. Experiments on MIMIC-III-full show that our ensemble learner substantially improves both macro and micro F1, from 10.4 to 14.6 and from 58.2 to 59.1, respectively.

Value estimates at multiple timescales can help create advanced discounting functions and allow agents to form more effective predictive models of their environment. In this work, we investigate learning over multiple horizons concurrently for off-policy reinforcement learning by using an advantage-based action selection method and introducing architectural improvements. Our proposed agent learns over multiple horizons simultaneously, while using either exponential or hyperbolic discounting functions. We implement our approach on Rainbow, a value-based off-policy algorithm, and test on Procgen, a collection of procedurally-generated environments, to demonstrate the effectiveness of this approach, specifically to evaluate the agent's performance in previously unseen scenarios.

Graph meta learning aims to learn historical knowledge from training graph neural networks (GNNs) models and adapt it to downstream learning tasks in a target graph, which has drawn increasing attention due to its ability of knowledge transfer and fast adaptation. While existing graph meta learning approaches assume the learning tasks are from the same graph domain but lack the solution for multi-domain adaptation. In this paper, we address the multi-domain generalized graph meta learning problem, which is challenging due to non-Euclidean data, inequivalent feature spaces, and heterogeneous distributions. To this end, we propose a novel solution called MD-Gram for multi-domain graph generalization. It introduces an empirical graph generalization method that uses empirical vectors to form a unified expression of non-Euclidean graph data. Then it proposes a multi-domain graphs transformation approach to transform the learning tasks from multiple source-domain graphs with inequivalent feature spaces into a common domain, where graph meta learning is conducted to learn generalized knowledge. It further adopts a domain-specific GNN enhancement method to learn a customized GNN model to achieve fast adaptation in the unseen target domain. Extensive experiments based on four real-world graph domain datasets show that the proposed method significantly outperforms the state-of-the-art in multi-domain graph meta learning tasks.

Active learning (AL) aims to find a better trade-off between labeling costs and model performance by consciously selecting more informative samples to label. Recently, adversarial approaches have emerged as effective solutions. Most of them leverage generative adversarial networks to align feature distributions of labeled and unlabeled data, upon which discriminators are trained to better distinguish between them. However, these methods fail to consider the relationship between unlabeled samples and decision boundaries, and their training processes are often complex and unstable. To this end, this paper proposes a novel adversarial AL method, namely multi-classifier adversarial optimization for active learning (MAOAL). MAOAL employs task-specific decision boundaries for data alignment while selecting the most informative samples to label. To fulfill this, we introduce a novel classifier class confusion (C3) metric, which represents the classifier discrepancy as the inter-class correlation of classifier outputs. Without any additional hyper-parameters, the C3 metric further reduces the negative impacts of ambiguous samples in the process of distribution alignment and sample selection. More concretely, the network is trained adversarially by adding two auxiliary classifiers, reducing the distribution bias of labeled and unlabeled samples by minimizing the C3 loss between classifiers, while learning tighter decision boundaries and highlighting hard samples by maximizing the C3 loss. Finally, the unlabeled samples with the highest C3 loss are selected to label. Extensive experiments demonstrate the superiority of our approach over state-of-the-art AL methods in terms of image classification and object detection.

Recent studies on knowledge graphs (KGs) show that path-based methods empowered by pre-trained language models perform well in the provision of inductive and explainable relation predictions. In this paper, we introduce the concepts of relation path coverage and relation path confidence to filter out unreliable paths prior to model training to elevate the model performance. Moreover, we propose Knowledge Reasoning Sentence Transformer (KRST) to predict inductive relations in KGs. KRST is designed to encode the extracted reliable paths in KGs, allowing us to properly cluster paths and provide multi-aspect explanations. We conduct extensive experiments on three real-world datasets. The experimental results show that compared to SOTA models, KRST achieves the best performance in most transductive and inductive test cases (4 of 6), and in 11 of 12 few-shot test cases.

Multi-action dialog policy (MADP), which generates multiple atomic dialog actions per turn, has been widely applied in task-oriented dialog systems to provide expressive and efficient system responses. Existing MADP models usually imitate action combinations from the labeled multi-action dialog samples. Due to data limitations, they generalize poorly toward unseen dialog flows. While reinforcement learning-based methods are proposed to incorporate the service ratings from real users and user simulators as external supervision signals, they suffer from sparse and less credible dialog-level rewards. To cope with this problem, we explore to improve MADPL with explicit and implicit turn-level user feedback received for historical predictions (i.e., logged user feedback) that are cost-efficient to collect and faithful to real-world scenarios. The task is challenging since the logged user feedback provides only partial label feedback limited to the particular historical dialog actions predicted by the agent. To fully exploit such feedback information, we propose BanditMatch, which addresses the task from a feedback-enhanced semi-supervised learning perspective with a hybrid learning objective of SSL and bandit learning. BanditMatch integrates pseudo-labeling methods to better explore the action space through constructing full label feedback. Extensive experiments show that our BanditMatch improves MADPL over the state-of-the-art methods by generating more concise and informative responses. The source code and the appendix of this paper can be obtained from https://github.com/ShuoZhangXJTU/BanditMatch.

Whole slide image (WSI) has been widely used to assist automated diagnosis under the deep learning fields. However, most previous works only discuss the SINGLE task setting which is not aligned with real clinical setting, where pathologists often conduct multiple diagnosis tasks simultaneously. Also, it is commonly recognized that the multi-task learning paradigm can improve learning efficiency by exploiting commonalities and differences across multiple tasks. To this end, we present a novel multi-task framework (i.e., MulGT) for WSI analysis by the specially designed Graph-Transformer equipped with Task-aware Knowledge Injection and Domain Knowledge-driven Graph Pooling modules. Basically, with the Graph Neural Network and Transformer as the building commons, our framework is able to learn task-agnostic low-level local information as well as task-specific high-level global representation. Considering that different tasks in WSI analysis depend on different features and properties, we also design a novel Task-aware Knowledge Injection module to transfer the task-shared graph embedding into task-specific feature spaces to learn more accurate representation for different tasks. Further, we elaborately design a novel Domain Knowledge-driven Graph Pooling module for each task to improve both the accuracy and robustness of different tasks by leveraging different diagnosis patterns of multiple tasks. We evaluated our method on two public WSI datasets from TCGA projects, i.e., esophageal carcinoma and kidney carcinoma. Experimental results show that our method outperforms single-task counterparts and the state-of-theart methods on both tumor typing and staging tasks.

Multi-label image classification is a foundational topic in various domains. Multimodal learning approaches have recently achieved outstanding results in image representation and single-label image classification. For instance, Contrastive Language-Image Pretraining (CLIP) demonstrates impressive image-text representation learning abilities and is robust to natural distribution shifts. This success inspires us to leverage multimodal learning for multi-label classification tasks, and benefit from contrastively learnt pretrained models. We propose the Multimodal Multi-label Image Classification (MuMIC) framework, which utilizes a hardness-aware tempered sigmoid based Binary Cross Entropy loss function, thus enables the optimization on multi-label objectives and transfer learning on CLIP. MuMIC is capable of providing high classification performance, handling real-world noisy data, supporting zero-shot predictions, and producing domain-specific image embeddings. In this study, a total of 120 image classes are defined, and more than 140K positive annotations are collected on approximately 60K Booking.com images. The final MuMIC model is deployed on Booking.com Content Intelligence Platform, and it outperforms other state-of-the-art models with 85.6% GAP@10 and 83.8% GAP on all 120 classes, as well as a 90.1% macro mAP score across 32 majority classes. We summarize the modelling choices which are extensively tested through ablation studies. To the best of our knowledge, we are the first to adapt contrastively learnt multimodal pretraining for real-world multi-label image classification problems, and the innovation can be transferred to other domains.

This paper considers the problem of cooperative localization of multiple robots under uncertainty, communicating over a partially connected, dynamic communication network and assisted by an agile landmark. Each robot owns an IMU and a relative pose sensing suite, which can get faulty due to system or environmental uncertainty, and therefore exhibit large bias in their estimation output. For the robots to localize accurately under sensor failure and system or environmental uncertainty, a novel Distributed Learning based Decentralized Cooperative Localization (DL-DCL) algorithm is proposed that involves real-time learning of an information fusion strategy by each robot for combining pose estimates from its own sensors as well as from those of its neighboring robots, and utilizing the moving landmark's pose information as a feedback to the learning process. Convergence analysis shows that the learning process converges exponentially under certain reasonable assumptions. Simulations involving sensor failures inducing around 40-60 times increase in the nominal bias show DL-DCL's estimation performance to be approximately 40% better than the well-known covariance-based estimate fusion methods. For the evaluation of DL-DCL's implementability and fault-tolerance capability in practice, a high-fidelity simulation is carried out in Gazebo with ROS2.

With artificial intelligence systems increasingly applied in consequential domains, researchers have begun to ask how AI systems ought to act in ethically charged situations where even humans lack consensus. In the Moral Machine project, researchers crowdsourced answers to "Trolley Problems" concerning autonomous vehicles. Subsequently, Noothigattu et al. (2018) proposed inferring linear functions that approximate each individual's preferences and aggregating these linear models by averaging parameters across the population. In this paper, we examine this averaging mechanism, focusing on fairness concerns and strategic effects. We investigate a simple setting where the population consists of two groups, the minority constitutes an α < 0.5 share of the population, and within-group preferences are homogeneous. Focusing on the fraction of contested cases where the minority group prevails, we make the following observations: (a) even when all parties report their preferences truthfully, the fraction of disputes where the minority prevails is less than proportionate in α; (b) the degree of sub-proportionality grows more severe as the level of disagreement between the groups increases; (c) when parties report preferences strategically, pure strategy equilibria do not always exist; and (d) whenever a pure strategy equilibrium exists, the majority group prevails 100% of the time. These findings raise concerns about stability and fairness of averaging as a mechanism for aggregating diverging voices. Finally, we discuss alternatives, including randomized dictatorship and median-based mechanisms.

Many real-world sequential decision problems can be decomposed into processes with independent dynamics that are coupled via the action structure. We discuss recent work on such problems and future directions.

Monitoring machine learning models once they are deployed is challenging. It is even more challenging to decide when to retrain models in real-case scenarios when labeled data is beyond reach, and monitoring performance metrics becomes unfeasible. In this work, we use non-parametric bootstrapped uncertainty estimates and SHAP values to provide explainable uncertainty estimation as a technique that aims to monitor the deterioration of machine learning models in deployment environments, as well as determine the source of model deteri- oration when target labels are not available. Classical methods are purely aimed at detecting distribution shift, which can lead to false positives in the sense that the model has not deterio- rated despite a shift in the data distribution. To estimate model uncertainty we construct prediction intervals using a novel bootstrap method, which improves previous state-of-the-art work. We show that both our model deterioration detection system as well as our uncertainty estimation method achieve better performance than the current state-of-the-art. Finally, we use explainable AI techniques to gain an understanding of the drivers of model deterioration. We release an open source Python package, doubt, which implements our pro- posed methods, as well as the code used to reproduce our experiments.

We study monitoring of linear-time arithmetic properties against finite traces generated by an unknown dynamic system. The monitoring state is determined by considering at once the trace prefix seen so far, and all its possible finite-length, future continuations. This makes monitoring at least as hard as satisfiability and validity. Traces consist of finite sequences of assignments of a fixed set of variables to numerical values. Properties are specified in a logic we call ALTLf, combining LTLf (LTL on finite traces) with linear arithmetic constraints that may carry lookahead, i.e., variables may be compared over multiple instants of the trace. While the monitoring problem for this setting is undecidable in general, we show decidability for (a) properties without lookahead, and (b) properties with lookahead that satisfy the abstract, semantic condition of finite summary, studied before in the context of model checking. We then single out concrete, practically relevant classes of constraints guaranteeing finite summary. Feasibility is witnessed by a prototype implementation.

Procuring expressive molecular representations underpins AI-driven molecule design and scientific discovery. The research mainly focuses on atom-level homogeneous molecular graphs, ignoring the rich information in subgraphs or motifs. However, it has been widely accepted that substructures play a dominant role in identifying and determining molecular properties. To address such issues, we formulate heterogeneous molecular graphs (HMGs) and introduce a novel architecture to exploit both molecular motifs and 3D geometry. Precisely, we extract functional groups as motifs for small molecules and employ reinforcement learning to adaptively select quaternary amino acids as motif candidates for proteins. Then HMGs are constructed with both atom-level and motif-level nodes. To better accommodate those HMGs, we introduce a variant of the Transformer named Molformer, which adopts a heterogeneous self-attention layer to distinguish the interactions between multi-level nodes. Besides, it is also coupled with a multi-scale mechanism to capture fine-grained local patterns with increasing contextual scales. An attentive farthest point sampling algorithm is also proposed to obtain the molecular representations. We validate Molformer across a broad range of domains, including quantum chemistry, physiology, and biophysics. Extensive experiments show that Molformer outperforms or achieves the comparable performance of several state-of-the-art baselines. Our work provides a promising way to utilize informative motifs from the perspective of multi-level graph construction. The code is available at https://github.com/smiles724/Molformer.

Research in model-based reinforcement learning has made significant progress in recent years. Compared to single-agent settings, the exponential dimension growth of the joint state-action space in multi-agent systems dramatically increases the complexity of the environment dynamics, which makes it infeasible to learn an accurate global model and thus necessitates the use of agent-wise local models. However, during multi-step model rollouts, the prediction of one local model can affect the predictions of other local models in the next step. As a result, local prediction errors can be propagated to other localities and eventually give rise to considerably large global errors. Furthermore, since the models are generally used to predict for multiple steps, simply minimizing one-step prediction errors regardless of their long-term effect on other models may further aggravate the propagation of local errors. To this end, we propose Models as AGents (MAG), a multi-agent model optimization framework that reversely treats the local models as multi-step decision making agents and the current policies as the dynamics during the model rollout process. In this way, the local models are able to consider the multi-step mutual affect between each other before making predictions. Theoretically, we show that the objective of MAG is approximately equivalent to maximizing a lower bound of the true environment return. Experiments on the challenging StarCraft II benchmark demonstrate the effectiveness of MAG.

When faced with novel tasks, humans have the ability to form successful strategies, seemingly without much effort. Artificial systems, on the other, hand cannot, at least when the flexibility at which humans perform is considered. For my dissertation, I am using program synthesis as a tool to study the factors that affect strategy choices in intelligent systems. I am evaluating my work through agents that reason through problems from the Abstract Reasoning Corpus and The Block Design Task.

Data Science (DS) is an interdisciplinary topic that is applicable to many domains. In this preliminary investigation, we use caselet, a mini-version of a case study, as a learning tool to allow students to practice data science problem solving (DSPS). Using a dataset collected from a real-world classroom, we performed correlation analysis to reveal the structure of cognition and metacognition processes. We also explored the similarity of different DS knowledge components based on students’ performance. In addition, we built a predictive model to characterize the relationship between metacognition, cognition, and learning gain.

The increased integration of artificial intelligence (AI) technologies in human workflows has resulted in a new paradigm of AI-assisted decision making, in which an AI model provides decision recommendations while humans make the final decisions. To best support humans in decision making, it is critical to obtain a quantitative understanding of how humans interact with and rely on AI. Previous studies often model humans' reliance on AI as an analytical process, i.e., reliance decisions are made based on cost-benefit analysis. However, theoretical models in psychology suggest that the reliance decisions can often be driven by emotions like humans' trust in AI models. In this paper, we propose a hidden Markov model to capture the affective process underlying the human-AI interaction in AI-assisted decision making, by characterizing how decision makers adjust their trust in AI over time and make reliance decisions based on their trust. Evaluations on real human behavior data collected from human-subject experiments show that the proposed model outperforms various baselines in accurately predicting humans' reliance behavior in AI-assisted decision making. Based on the proposed model, we further provide insights into how humans' trust and reliance dynamics in AI-assisted decision making is influenced by contextual factors like decision stakes and their interaction experiences.

We investigate the complexity of the model-checking problem for a family of modal logics capturing the notion of “knowing how”. We consider the most standard ability-based knowing how logic, for which we show that model-checking is PSpace-complete. By contrast, a multi-agent variant based on an uncertainty relation between plans in which uncertainty is encoded by a regular language, is shown to admit a PTime model-checking problem. We extend with budgets the above-mentioned ability-logics, as done for ATL-like logics. We show that for the former logic enriched with budgets, the complexity increases to at least ExpSpace-hardness, whereas for the latter, the PTime bound is preserved. Other variant logics are discussed along the paper.