AAAI 2023 论文列表

Accurately predicting the volume of amniotic fluid is fundamental to assessing pregnancy risks, though the task usually requires many hours of laborious work by medical experts. In this paper, we present AmnioML, a machine learning solution that leverages deep learning and conformal prediction to output fast and accurate volume estimates and segmentation masks from fetal MRIs with Dice coefficient over 0.9. Also, we make available a novel, curated dataset for fetal MRIs with 853 exams and benchmark the performance of many recent deep learning architectures. In addition, we introduce a conformal prediction tool that yields narrow predictive intervals with theoretically guaranteed coverage, thus aiding doctors in detecting pregnancy risks and saving lives. A successful case study of AmnioML deployed in a medical setting is also reported. Real-world clinical benefits include up to 20x segmentation time reduction, with most segmentations deemed by doctors as not needing any further manual refinement. Furthermore, AmnioML's volume predictions were found to be highly accurate in practice, with mean absolute error below 56mL and tight predictive intervals, showcasing its impact in reducing pregnancy complications.

Variational quantum algorithms (VQAs) are the quantum analog of classical neural networks (NNs). A VQA consists of a parameterized quantum circuit (PQC) which is composed of multiple layers of ansatzes (simpler PQCs, which are an analogy of NN layers) that differ only in selections of parameters. Previous work has identified the alternating layered ansatz as potentially a new standard ansatz in near-term quantum computing. Indeed, shallow alternating layered VQAs are easy to implement and have been shown to be both trainable and expressive. In this work, we introduce a training algorithm with an exponential reduction in training cost of such VQAs. Moreover, our algorithm uses classical shadows of quantum input data, and can hence be run on a classical computer with rigorous performance guarantees. We demonstrate 2-3 orders of magnitude improvement in the training cost using our algorithm for the example problems of finding state preparation circuits and the quantum autoencoder.

Reinforcement learning has been used to approach well-known NP-hard combinatorial problems in graph theory. Among these, Hamiltonian cycle problems are exceptionally difficult to analyze, even when restricted to individual instances of structurally complex graphs. In this paper, we use Monte Carlo Tree Search (MCTS), the search algorithm behind many state-of-the-art reinforcement learning algorithms such as AlphaZero, to create autonomous agents that learn to play the game of Snake, a game centered on properties of Hamiltonian cycles on grid graphs. The game of Snake can be formulated as a single-player discounted Markov Decision Process (MDP), where the agent must behave optimally in a stochastic environment. Determining the optimal policy for Snake, defined as the policy that maximizes the probability of winning -- or win rate -- with higher priority and minimizes the expected number of time steps to win with lower priority, is conjectured to be NP-hard. Performance-wise, compared to prior work in the Snake game, our algorithm is the first to achieve a win rate over 0.5 (a uniform random policy achieves a win rate < 2.57 x 10^{-15}), demonstrating the versatility of AlphaZero in tackling NP-hard problems.

This paper proposes AlphaRoute, an AlphaGo inspired algorithm for coordinating large-scale routes, built upon graph attention reinforcement learning and Monte Carlo Tree Search (MCTS). We first partition the road network into regions and model large-scale coordinated route planning as a Markov game, where each partitioned region is treated as a player instead of each driver. Then, AlphaRoute applies a bilevel optimization framework, consisting of several region planners and a global planner, where the region planner coordinates the route choices for vehicles located in the region and generates several strategies, and the global planner evaluates the combination of strategies. AlphaRoute is built on graph attention network for evaluating each state and MCTS algorithm for dynamically visiting and simulating the future state for narrowing down the search space. AlphaRoute is capable of 1) bridging user fairness and system efficiency, 2) achieving higher search efficiency by alleviating the curse of dimensionality problems, and 3) making an effective and informed route planning by simulating over the future to capture traffic dynamics. Comprehensive experiments are conducted on two real-world road networks as compared with several baselines to evaluate the performance, and results show that AlphaRoute achieves the lowest travel time, and is efficient and effective for coordinating large-scale routes and alleviating the traffic congestion problem. The code will be publicly available.

We study the problem of best arm identification in a federated learning multi-armed bandit setup with a central server and multiple clients. Each client is associated with a multi-armed bandit in which each arm yields i.i.d. rewards following a Gaussian distribution with an unknown mean and known variance. The set of arms is assumed to be the same at all the clients. We define two notions of best arm local and global. The local best arm at a client is the arm with the largest mean among the arms local to the client, whereas the global best arm is the arm with the largest average mean across all the clients. We assume that each client can only observe the rewards from its local arms and thereby estimate its local best arm. The clients communicate with a central server on uplinks that entail a cost of C>=0 units per usage per uplink. The global best arm is estimated at the server. The goal is to identify the local best arms and the global best arm with minimal total cost, defined as the sum of the total number of arm selections at all the clients and the total communication cost, subject to an upper bound on the error probability. We propose a novel algorithm FedElim that is based on successive elimination and communicates only in exponential time steps and obtain a high probability instance-dependent upper bound on its total cost. The key takeaway from our paper is that for any C>=0 and error probabilities sufficiently small, the total number of arm selections (resp. the total cost) under FedElim is at most 2 (resp. 3) times the maximum total number of arm selections under its variant that communicates in every time step. Additionally, we show that the latter is optimal in expectation up to a constant factor, thereby demonstrating that communication is almost cost-free in FedElim. We numerically validate the efficacy of FedElim on two synthetic datasets and the MovieLens dataset.

Alignment between image and text has shown promising improvements on patch-level pre-trained document image models. However, investigating more effective or finer-grained alignment techniques during pre-training requires a large amount of computation cost and time. Thus, a question naturally arises: Could we fine-tune the pre-trained models adaptive to downstream tasks with alignment objectives and achieve comparable or better performance? In this paper, we propose a new model architecture with alignment-enriched tuning (dubbed AETNet) upon pre-trained document image models, to adapt downstream tasks with the joint task-specific supervised and alignment-aware contrastive objective. Specifically, we introduce an extra visual transformer as the alignment-ware image encoder and an extra text transformer as the alignment-ware text encoder before multimodal fusion. We consider alignment in the following three aspects: 1) document-level alignment by leveraging the cross-modal and intra-modal contrastive loss; 2) global-local alignment for modeling localized and structural information in document images; and 3) local-level alignment for more accurate patch-level information. Experiments on various downstream tasks show that AETNet can achieve state-of-the-art performance on various downstream tasks. Notably, AETNet consistently outperforms state-of-the-art pre-trained models, such as LayoutLMv3 with fine-tuning techniques, on three different downstream tasks. Code is available at https://github.com/MAEHCM/AET.

Air pollution is a crucial issue affecting human health and livelihoods, as well as one of the barriers to economic growth. Forecasting air quality has become an increasingly important endeavor with significant social impacts, especially in emerging countries. In this paper, we present a novel Transformer termed AirFormer to predict nationwide air quality in China, with an unprecedented fine spatial granularity covering thousands of locations. AirFormer decouples the learning process into two stages: 1) a bottom-up deterministic stage that contains two new types of self-attention mechanisms to efficiently learn spatio-temporal representations; 2) a top-down stochastic stage with latent variables to capture the intrinsic uncertainty of air quality data. We evaluate AirFormer with 4-year data from 1,085 stations in Chinese Mainland. Compared to prior models, AirFormer reduces prediction errors by 5%∼8% on 72-hour future predictions. Our source code is available at https://github.com/yoshall/airformer.

Image aesthetic quality assessment (AQA) aims to assign numerical aesthetic ratings to images whilst image aesthetic captioning (IAC) aims to generate textual descriptions of the aesthetic aspects of images. In this paper, we study image AQA and IAC together and present a new IAC method termed Aesthetically Relevant Image Captioning (ARIC). Based on the observation that most textual comments of an image are about objects and their interactions rather than aspects of aesthetics, we first introduce the concept of Aesthetic Relevance Score (ARS) of a sentence and have developed a model to automatically label a sentence with its ARS. We then use the ARS to design the ARIC model which includes an ARS weighted IAC loss function and an ARS based diverse aesthetic caption selector (DACS). We present extensive experimental results to show the soundness of the ARS concept and the effectiveness of the ARIC model by demonstrating that texts with higher ARS’s can predict the aesthetic ratings more accurately and that the new ARIC model can generate more accurate, aesthetically more relevant and more diverse image captions. Furthermore, a large new research database containing 510K images with over 5 million comments and 350K aesthetic scores, and code for implementing ARIC, are available at https://github.com/PengZai/ARIC

Data augmentation is widely used in text classification, especially in the low-resource regime where a few examples for each class are available during training. Despite the success, generating data augmentations as hard positive examples that may increase their effectiveness is under-explored. This paper proposes an Adversarial Word Dilution (AWD) method that can generate hard positive examples as text data augmentations to train the low-resource text classification model efficiently. Our idea of augmenting the text data is to dilute the embedding of strong positive words by weighted mixing with unknown-word embedding, making the augmented inputs hard to be recognized as positive by the classification model. We adversarially learn the dilution weights through a constrained min-max optimization process with the guidance of the labels. Empirical studies on three benchmark datasets show that AWD can generate more effective data augmentations and outperform the state-of-the-art text data augmentation methods. The additional analysis demonstrates that the data augmentations generated by AWD are interpretable and can flexibly extend to new examples without further training.

A lot of theoretical and empirical evidence shows that the flatter local minima tend to improve generalization. Adversarial Weight Perturbation (AWP) is an emerging technique to efficiently and effectively find such minima. In AMP we minimize the loss w.r.t. a bounded worst-case perturbation of the model parameters thereby favoring local minima with a small loss in a neighborhood around them. The benefits of AWP, and more generally the connections between flatness and generalization, have been extensively studied for i.i.d. data such as images. In this paper, we extensively study this phenomenon for graph data. Along the way, we first derive a generalization bound for non-i.i.d. node classification tasks. Then we identify a vanishing-gradient issue with all existing formulations of AWP and we propose a new Weighted Truncated AWP (WT-AWP) to alleviate this issue. We show that regularizing graph neural networks with WT-AWP consistently improves both natural and robust generalization across many different graph learning tasks and models.

Deep neural models (e.g. Transformer) naturally learn spurious features, which create a ``shortcut'' between the labels and inputs, thus impairing the generalization and robustness. This paper advances self-attention mechanism to its robust variant for Transformer-based pre-trained language models (e.g. BERT). We propose Adversarial Self-Attention mechanism (ASA), which adversarially biases the attentions to effectively suppress the model reliance on features (e.g. specific keywords) and encourage its exploration of broader semantics. We conduct comprehensive evaluation across a wide range of tasks for both pre-training and fine-tuning stages. For pre-training, ASA unfolds remarkable performance gain compared to naive training for longer steps. For fine-tuning, ASA-empowered models outweigh naive models by a large margin considering both generalization and robustness.

Learning from raw high dimensional data via interaction with a given environment has been effectively achieved through the utilization of deep neural networks. Yet the observed degradation in policy performance caused by imperceptible worst-case policy dependent translations along high sensitivity directions (i.e. adversarial perturbations) raises concerns on the robustness of deep reinforcement learning policies. In our paper, we show that these high sensitivity directions do not lie only along particular worst-case directions, but rather are more abundant in the deep neural policy landscape and can be found via more natural means in a black-box setting. Furthermore, we show that vanilla training techniques intriguingly result in learning more robust policies compared to the policies learnt via the state-of-the-art adversarial training techniques. We believe our work lays out intriguing properties of the deep reinforcement learning policy manifold and our results can help to build robust and generalizable deep reinforcement learning policies.

Source-free object detection (SFOD) aims to transfer a detector pre-trained on a label-rich source domain to an unlabeled target domain without seeing source data. While most existing SFOD methods generate pseudo labels via a source-pretrained model to guide training, these pseudo labels usually contain high noises due to heavy domain discrepancy. In order to obtain better pseudo supervisions, we divide the target domain into source-similar and source-dissimilar parts and align them in the feature space by adversarial learning.Specifically, we design a detection variance-based criterion to divide the target domain. This criterion is motivated by a finding that larger detection variances denote higher recall and larger similarity to the source domain. Then we incorporate an adversarial module into a mean teacher framework to drive the feature spaces of these two subsets indistinguishable. Extensive experiments on multiple cross-domain object detection datasets demonstrate that our proposed method consistently outperforms the compared SFOD methods. Our implementation is available at https://github.com/ChuQiaosong.

In the talk, I shall describe my lab’s recent advances in AI, applied machine learning, and data mining to combat malicious actors (sockpuppets, ban evaders, etc.) and dangerous content (misinformation, hate, etc.) on web and social media platforms. My vision is to create a trustworthy online ecosystem for everyone and create the next generation of socially-aware methods that promote health, equity, and safety. Broadly, in my research, I have created novel graph, content (NLP, multimodality), and adversarial machine learning methods leveraging terabytes of data to detect, predict, and mitigate online threats. I shall describe the advancements made in my group across four key thrusts: (1) Detection of harmful content and malicious actors across platforms, languages, and modalities, (2) Robustifying detection models against adversarial actors by predicting future malicious activities, (3) Attributing the impact of harmful content and the role of recommender systems, and (4) Developing mitigation techniques to counter misinformation by professionals and the crowd.

It is an open question what semantic representations transformer-based language models can encode and whether they have access to more abstract aspects of semantic meaning. Here, we propose a diagnostic dataset to investigate how well language models understand the degree semantics of adjectives. In the dataset, referred as the Adjective Scale Probe (ASP), we semi-automatically generate 8 tests of Natural Language Inference (NLI) questions to test 8 key capabilities of adjective interpretation. We apply the ASP dataset to evaluate the performance of 3 language models, i.e., BERT, DeBERTa, and T0. It is found that language models perform below the majority baseline for most tests of the ASP, even when the models have been fine-tuned to achieve high performance on the large-scale MNLI dataset. But after we fine-tune the pre-trained models on a subset of the ASP, DeBERTa can achieve high performance on the untrained adjectives and untrained tests, suggesting that DeBERTa may have captured degree semantic information of adjectives through pre-training but it needs specific training data to learn how to apply such information to the current tasks. In sum, the ASP provides an easy-to-use method to test fine-grained formal semantic properties of adjectives, and reveals language models' abilities to access formal semantic information.

Domain generalization in semantic segmentation aims to alleviate the performance degradation on unseen domains through learning domain-invariant features. Existing methods diversify images in the source domain by adding complex or even abnormal textures to reduce the sensitivity to domain-specific features. However, these approaches depends heavily on the richness of the texture bank and training them can be time-consuming. In contrast to importing textures arbitrarily or augmenting styles randomly, we focus on the single source domain itself to achieve the generalization. In this paper, we present a novel adaptive texture filtering mechanism to suppress the influence of texture without using augmentation, thus eliminating the interference of domain-specific features. Further, we design a hierarchical guidance generalization network equipped with structure-guided enhancement modules, which purpose to learn the domain-invariant generalized knowledge. Extensive experiments together with ablation studies on widely-used datasets are conducted to verify the effectiveness of the proposed model, and reveal its superiority over other state-of-the-art alternatives.

With the fast development of offshore wind farms as renewable energy sources, maintaining them efficiently and safely becomes necessary. The high costs of operation and maintenance (O&M) are due to the length of turbine downtime and the logistics for human technician transfer. To reduce such costs, we propose a comprehensive multi-robot system that includes unmanned aerial vehicles (UAV), autonomous surface vessels (ASV), and inspection-and-repair robots (IRR). Our system, which is capable of co-managing the farms with human operators located onshore, brings down costs and significantly reduces the Health and Safety (H&S) risks of O&M by assisting human operators in performing dangerous tasks. In this paper, we focus on using AI temporal planning to coordinate the actions of the different autonomous robots that form the multi-robot system. We devise a new, adaptive planning approach that reduces failures and replanning by performing data-driven goal and domain refinement. Our experiments in both simulated and real-world scenarios prove the effectiveness and robustness of our technique. The success of our system marks the first-step towards a large-scale, multirobot solution for wind farm O&M.

Conventional reinforcement learning (RL) needs an environment to collect fresh data, which is impractical when online interactions are costly. Offline RL provides an alternative solution by directly learning from the previously collected dataset. However, it will yield unsatisfactory performance if the quality of the offline datasets is poor. In this paper, we consider an offline-to-online setting where the agent is first learned from the offline dataset and then trained online, and propose a framework called Adaptive Policy Learning for effectively taking advantage of offline and online data. Specifically, we explicitly consider the difference between the online and offline data and apply an adaptive update scheme accordingly, that is, a pessimistic update strategy for the offline dataset and an optimistic/greedy update scheme for the online dataset. Such a simple and effective method provides a way to mix the offline and online RL and achieve the best of both worlds. We further provide two detailed algorithms for implementing the framework through embedding value or policy-based RL algorithms into it. Finally, we conduct extensive experiments on popular continuous control tasks, and results show that our algorithm can learn the expert policy with high sample efficiency even when the quality of offline dataset is poor, e.g., random dataset.

The integration of discrete algorithmic components in deep learning architectures has numerous applications. Recently, Implicit Maximum Likelihood Estimation, a class of gradient estimators for discrete exponential family distributions, was proposed by combining implicit differentiation through perturbation with the path-wise gradient estimator. However, due to the finite difference approximation of the gradients, it is especially sensitive to the choice of the finite difference step size, which needs to be specified by the user. In this work, we present Adaptive IMLE (AIMLE), the first adaptive gradient estimator for complex discrete distributions: it adaptively identifies the target distribution for IMLE by trading off the density of gradient information with the degree of bias in the gradient estimates. We empirically evaluate our estimator on synthetic examples, as well as on Learning to Explain, Discrete Variational Auto-Encoders, and Neural Relational Inference tasks. In our experiments, we show that our adaptive gradient estimator can produce faithful estimates while requiring orders of magnitude fewer samples than other gradient estimators.