AAAI 2024 Papers

In the setting of conference peer review, the conference aims to accept high-quality papers and reject low-quality papers based on noisy review scores. A recent work proposes the isotonic mechanism, which can elicit the ranking of paper qualities from an author with multiple submissions to help improve the conference's decisions. However, the isotonic mechanism relies on the assumption that the author's utility is both an increasing and a convex function with respect to the review score, which is often violated in realistic settings (e.g.~when authors aim to maximize the number of accepted papers). In this paper, we propose a sequential review mechanism that can truthfully elicit the ranking information from authors while only assuming the agent's utility is increasing with respect to the true quality of her accepted papers. The key idea is to review the papers of an author in a sequence based on the provided ranking and conditioning the review of the next paper on the review scores of the previous papers. Advantages of the sequential review mechanism include: 1) eliciting truthful ranking information in a more realistic setting than prior work; 2) reducing the reviewing workload and increasing the average quality of papers being reviewed; 3) incentivizing authors to write fewer papers of higher quality.

Automatic mitochondrial segmentation enjoys great popularity with the development of deep learning. However, the coarse prediction raised by the presence of regular 3D grids in previous methods regardless of 3D CNN or the vision transformers suggest a possibly sub-optimal feature arrangement. To mitigate this limitation, we attempt to interpret the 3D EM image stacks as a set of interrelated 3D fragments for a better solution. However, it is non-trivial to model the 3D fragments without introducing excessive computational overhead. In this paper, we design a coherent fragment vision transformer (FragViT) combined with affinity learning to manipulate features on 3D fragments yet explore mutual relationships to model fragment-wise context, enjoying locality prior without sacrificing global reception. The proposed FragViT includes a fragment encoder and a hierarchical fragment aggregation module. The fragment encoder is equipped with affinity heads to transform the tokens into fragments with homogeneous semantics, and the multi-layer self-attention is used to explicitly learn inter-fragment relations with long-range dependencies. The hierarchical fragment aggregation module is responsible for hierarchically aggregating fragment-wise prediction back to the final voxel-wise prediction in a progressive manner. Extensive experimental results on the challenging MitoEM, Lucchi, and AC3/AC4 benchmarks demonstrate the effectiveness of the proposed method.

Toxic content detection is crucial for online services to remove inappropriate content that violates community standards. To automate the detection process, prior works have proposed varieties of machine learning (ML) approaches to train Language Models (LMs) for toxic content detection. However, both their accuracy and transferability across datasets are limited. Recently, Large Language Models (LLMs) have shown promise in toxic content detection due to their superior zero-shot and few-shot in-context learning ability as well as broad transferability on ML tasks. However, efficiently designing prompts for LLMs remains challenging. Moreover, the high run-time cost of LLMs may hinder their deployments in production. To address these challenges, in this work, we propose BD-LLM, a novel and efficient approach to bootstrapping and distilling LLMs for toxic content detection. Specifically, we design a novel prompting method named Decision-Tree-of-Thought (DToT) to bootstrap LLMs' detection performance and extract high-quality rationales. DToT can automatically select more fine-grained context to re-prompt LLMs when their responses lack confidence. Additionally, we use the rationales extracted via DToT to fine-tune student LMs. Our experimental results on various datasets demonstrate that DToT can improve the accuracy of LLMs by up to 4.6%. Furthermore, student LMs fine-tuned with rationales extracted via DToT outperform baselines on all datasets with up to 16.9% accuracy improvement, while being more than 60x smaller than conventional LLMs. Finally, we observe that student LMs fine-tuned with rationales exhibit better cross-dataset transferability.

Exploring biologically plausible algorithms as alternatives to error backpropagation (BP) is a challenging research topic in artificial intelligence. It also provides insights into the brain's learning methods. Recently, when combined with well-designed feedback loss functions such as Local Difference Reconstruction Loss (LDRL) and through hierarchical training of feedback pathway synaptic weights, Target Propagation (TP) has achieved performance comparable to BP in image classification tasks. However, with an increase in the number of network layers, the tuning and training cost of feedback weights escalates. Drawing inspiration from the work of Ernoult et al., we propose a training method that seeks the optimal solution for feedback weights. This method enhances the efficiency of feedback training by analytically minimizing feedback loss, allowing the feedback layer to skip certain local training iterations. More specifically, we introduce the Jacobian matching loss (JML) for feedback training. We also proactively implement layers designed to derive analytical solutions that minimize JML. Through experiments, we have validated the effectiveness of this approach. Using the CIFAR-10 dataset, our method showcases accuracy levels comparable to state-of-the-art TP methods. Furthermore, we have explored its effectiveness in more intricate network architectures.

The next generation of machine intelligence requires the capability of continual learning to acquire new knowledge without forgetting the old one while conserving limited computing resources. Spiking neural networks (SNNs), compared to artificial neural networks (ANNs), have more characteristics that align with biological neurons, which may be helpful as a potential gating function for knowledge maintenance in neural networks. Inspired by the selective sparse activation principle of context gating in biological systems, we present a novel SNN model with selective activation to achieve continual learning. The trace-based K-Winner-Take-All (K-WTA) and variable threshold components are designed to form the sparsity in selective activation in spatial and temporal dimensions of spiking neurons, which promotes the subpopulation of neuron activation to perform specific tasks. As a result, continual learning can be maintained by routing different tasks via different populations of neurons in the network. The experiments are conducted on MNIST and CIFAR10 datasets under the class incremental setting. The results show that the proposed SNN model achieves competitive performance similar to and even surpasses the other regularization-based methods deployed under traditional ANNs.

Satellite data bolstered by their increasing accessibility is leading to many endeavors of automated monitoring of the earth's surface for various applications. Such applications demand high spatial resolution images at a temporal resolution of a few days which entails the challenge of processing a huge volume of image time series data. To overcome this computing bottleneck, we present PatchNet, a bespoke adaptation of beam search and attention mechanism. PatchNet is an automated patch selection neural network that requires only a partial spatial traversal of an image time series and yet achieves impressive results. Satellite systems face a trade-off between spatial and temporal resolutions due to budget/technical constraints e.g., Landsat-8/9 or Sentinel-2 have high spatial resolution whereas, MODIS has high temporal resolution. To deal with the limitation of coarse temporal resolution, we propose FuSITSNet, a twofold feature-based generic fusion model with multimodal learning in a contrastive setting. It produces a learned representation after fusion of two satellite image time series leveraging finer spatial resolution of Landsat and finer temporal resolution of MODIS. The patch alignment module of FuSITSNet aligns the PatchNet processed patches of Landsat-8 with the corresponding MODIS regions to incorporate its finer resolution temporal features. The untraversed patches are handled by the cross-modality attention which highlights additional hot spot features from the two modalities. We conduct extensive experiments on more than 2000 counties of US for crop yield, snow cover, and solar energy prediction and show that even one-fourth spatial processing of image time series produces state-of-the-art results. FuSITSNet outperforms the predictions of single modality and data obtained using existing generative fusion models and allows for monitoring of dynamic phenomena using freely accessible images, thereby unlocking new opportunities.

Crowdsourcing platforms use various truth discovery algorithms to aggregate annotations from multiple labelers. In an online setting, however, the main challenge is to decide whether to ask for more annotations for each item to efficiently trade off cost (i.e., the number of annotations) for quality of the aggregated annotations. In this paper, we propose a novel approach for general complex annotation (such as bounding boxes and taxonomy paths), that works in an online crowdsourcing setting. We prove that the expected average similarity of a labeler is linear in their accuracy conditional on the reported label. This enables us to infer reported label accuracy in a broad range of scenarios. We conduct extensive evaluations on real-world crowdsourcing data from Meta and show the effectiveness of our proposed online algorithms in improving the cost-quality trade-off.

In numerous applications, binary reactions or event counts are observed and stored within high-order tensors. Tensor decompositions (TDs) serve as a powerful tool to handle such high-dimensional and sparse data. However, many traditional TDs are explicitly or implicitly designed based on the Gaussian distribution, which is unsuitable for discrete data. Moreover, most TDs rely on predefined multi-linear structures, such as CP and Tucker formats. Therefore, they may not be effective enough to handle complex real-world datasets. To address these issues, we propose ENTED, an Efficient Nonparametric TEnsor Decomposition for binary and count tensors. Specifically, we first employ a nonparametric Gaussian process (GP) to replace traditional multi-linear structures. Next, we utilize the Pólya-Gamma augmentation which provides a unified framework to establish conjugate models for binary and count distributions. Finally, to address the computational issue of GPs, we enhance the model by incorporating sparse orthogonal variational inference of inducing points, which offers a more effective covariance approximation within GPs and stochastic natural gradient updates for nonparametric models. We evaluate our model on several real-world tensor completion tasks, considering binary and count datasets. The results manifest both better performance and computational advantages of the proposed model.

The look-up table (LUT) has recently shown its practicability and effectiveness in super-resolution (SR) tasks due to its low computational cost and hardware independence. However, most existing methods focus on improving the performance of SR, neglecting the demand for high-speed SR on low-computational edge devices. In this paper, we propose an efficient expanded convolution (EC) layer, which expands the output size of regular convolution to enlarge the receptive field (RF) indirectly. It can increase the size of the LUT corresponding to the network linearly with the increase of RF. Additionally, after introducing the EC, multiple LUTs are merged into one LUT, achieving faster running speed while maintaining SR performance. More specifically, we expand the coverage of the convolutional output so that the output at the current position covers the target position and its surroundings, forming an overlapping sliding window at the output end. We sum up the overlapping parts of the sliding window as the output, thereby achieving the effect of enlarging the RF size. Moreover, by expanding the numerical range of the accumulated results and rescaling them to [0,255], the method can mitigate the error caused by quantization output. Experiments indicate that the proposed method performs better than the baseline method and is faster than other LUT-based SR methods.

Transformer has shown outstanding performance on image denoising, but the existing Transformer methods for image denoising are with large model sizes and high computational complexity, which is unfriendly to resource-constrained devices. In this paper, we propose a Lightweight Image Denoising Transformer method (LIDFormer) based on Triple Multi-Dconv Head Transposed Attention (TMDTA) to boost computational efficiency. LIDFormer first implements Discrete Wavelet Transform (DWT), which transforms the input image into a low-frequency space, greatly reducing the computational complexity of image denoising. However, the low-frequency image lacks fine-feature information, which degrades the denoising performance. To handle this problem, we introduce the Complementary Periodic Feature Reusing (CPFR) scheme for aggregating the shallow-layer features and the deep-layer features. Furthermore, TMDTA is proposed to integrate global context along three dimensions, thereby enhancing the ability of global feature representation. Note that our method can be applied as a pipeline for both convolutional neural networks and Transformers. Extensive experiments on several benchmarks demonstrate that the proposed LIDFormer achieves a better trade-off between high performance and low computational complexity on real-world image denoising tasks.

Accelerating the learning of Partial Differential Equations (PDEs) from experimental data will speed up the pace of scientific discovery. Previous randomized algorithms exploit sparsity in PDE updates for acceleration. However such methods are applicable to a limited class of decomposable PDEs, which have sparse features in the value domain. We propose Reel, which accelerates the learning of PDEs via random projection and has much broader applicability. Reel exploits the sparsity by decomposing dense updates into sparse ones in both the value and frequency domains. This decomposition enables efficient learning when the source of the updates consists of gradually changing terms across large areas (sparse in the frequency domain) in addition to a few rapid updates concentrated in a small set of “interfacial” regions (sparse in the value domain). Random projection is then applied to compress the sparse signals for learning. To expand the model applicability, Taylor series expansion is used in Reel to approximate the nonlinear PDE updates with polynomials in the decomposable form. Theoretically, we derive a constant factor approximation between the projected loss function and the original one with poly-logarithmic number of projected dimensions. Experimentally, we provide empirical evidence that our proposed Reel can lead to faster learning of PDE models (70-98% reduction in training time when the data is compressed to 1% of its original size) with comparable quality as the non-compressed models.

We study online learning and equilibrium computation in games with polyhedral decision sets, a property shared by normal-form games (NFGs) and extensive-form games (EFGs), when the learning agent is restricted to utilizing a best-response oracle. We show how to achieve constant regret in zero-sum games and O(T^0.25) regret in general-sum games while using only O(log t) best-response queries at a given iteration t, thus improving over the best prior result, which required O(T) queries per iteration. Moreover, our framework yields the first last-iterate convergence guarantees for self-play with best-response oracles in zero-sum games. This convergence occurs at a linear rate, though with a condition-number dependence. We go on to show a O(T^(-0.5)) best-iterate convergence rate without such a dependence. Our results build on linear-rate convergence results for variants of the Frank-Wolfe (FW) algorithm for strongly convex and smooth minimization problems over polyhedral domains. These FW results depend on a condition number of the polytope, known as facial distance. In order to enable application to settings such as EFGs, we show two broad new results: 1) the facial distance for polytopes in standard form is at least γ/k where γ is the minimum value of a nonzero coordinate of a vertex of the polytope and k≤n is the number of tight inequality constraints in the optimal face, and 2) the facial distance for polytopes of the form Ax=b, Cx≤d, x≥0 where x∈R^n, C≥0 is a nonzero integral matrix, and d≥0, is at least 1/(c√n), where c is the infinity norm of C. This yields the first such results for several problems such as sequence-form polytopes, flow polytopes, and matching polytopes.

The Mixture-of-Experts (MoE) approach has demonstrated outstanding scalability in multi-task learning including low-level upstream tasks such as concurrent removal of multiple adverse weather effects. However, the conventional MoE architecture with parallel Feed Forward Network (FFN) experts leads to significant parameter and computational overheads that hinder its efficient deployment. In addition, the naive MoE linear router is suboptimal in assigning task-specific features to multiple experts which limits its further scalability. In this work, we propose an efficient MoE architecture with weight sharing across the experts. Inspired by the idea of linear feature modulation (FM), our architecture implicitly instantiates multiple experts via learnable activation modulations on a single shared expert block. The proposed Feature Modulated Expert (FME) serves as a building block for the novel Mixture-of-Feature-Modulation-Experts (MoFME) architecture, which can scale up the number of experts with low overhead. We further propose an Uncertainty-aware Router (UaR) to assign task-specific features to different FM modules with well-calibrated weights. This enables MoFME to effectively learn diverse expert functions for multiple tasks. The conducted experiments on the multi-deweather task show that our MoFME outperforms the state-of-the-art in the image restoration quality by 0.1-0.2 dB while saving more than 74% of parameters and 20% inference time over the conventional MoE counterpart. Experiments on the downstream segmentation and classification tasks further demonstrate the generalizability of MoFME to real open-world applications.

Current methods for solving Stochastic Shortest Path Problems (SSPs) find states’ costs-to-go by applying Bellman backups, where state-of-the-art methods employ heuristics to select states to back up and prune. A fundamental limitation of these algorithms is their need to compute the cost-to-go for every applicable action during each state backup, leading to unnecessary computation for actions identified as sub-optimal. We present new connections between planning and operations research and, using this framework, we address this issue of unnecessary computation by introducing an efficient version of constraint generation for SSPs. This technique allows algorithms to ignore sub-optimal actions and avoid computing their costs-to-go. We also apply our novel technique to iLAO* resulting in a new algorithm, CG-iLAO*. Our experiments show that CG-iLAO* ignores up to 57% of iLAO*’s actions and it solves problems up to 8x and 3x faster than LRTDP and iLAO*.

Center-based clustering has attracted significant research interest from both theory and practice. In many practical applications, input data often contain background knowledge that can be used to improve clustering results. In this work, we build on widely adopted k-center clustering and model its input background knowledge as must-link (ML) and cannot-link (CL) constraint sets. However, most clustering problems including k-center are inherently NP-hard, while the more complex constrained variants are known to suffer severer approximation and computation barriers that significantly limit their applicability. By employing a suite of techniques including reverse dominating sets, linear programming (LP) integral polyhedron, and LP duality, we arrive at the first efficient approximation algorithm for constrained k-center with the best possible ratio of 2. We also construct competitive baseline algorithms and empirically evaluate our approximation algorithm against them on a variety of real datasets. The results validate our theoretical findings and demonstrate the great advantages of our algorithm in terms of clustering cost, clustering quality, and running time.

Image super-resolution is a fundamentally ill-posed problem because multiple valid high-resolution images exist for one low-resolution image. Super-resolution methods based on diffusion probabilistic models can deal with the ill-posed nature by learning the distribution of high-resolution images conditioned on low-resolution images, avoiding the problem of blurry images in PSNR-oriented methods. However, existing diffusion-based super-resolution methods have high time consumption with the use of iterative sampling, while the quality and consistency of generated images are less than ideal due to problems like color shifting. In this paper, we propose Efficient Conditional Diffusion Model with Probability Flow Sampling (ECDP) for image super-resolution. To reduce the time consumption, we design a continuous-time conditional diffusion model for image super-resolution, which enables the use of probability flow sampling for efficient generation. Additionally, to improve the consistency of generated images, we propose a hybrid parametrization for the denoiser network, which interpolates between the data-predicting parametrization and the noise-predicting parametrization for different noise scales. Moreover, we design an image quality loss as a complement to the score matching loss of diffusion models, further improving the consistency and quality of super-resolution. Extensive experiments on DIV2K, ImageNet, and CelebA demonstrate that our method achieves higher super-resolution quality than existing diffusion-based image super-resolution methods while having lower time consumption. Our code is available at https://github.com/Yuan-Yutao/ECDP.

We present an FCA-based axiomatization method that produces a complete OWL 2 EL TBox (the terminological part of an OWL 2 EL ontology) from a graph dataset in at most exponential time. We describe technical details that allow for efficient implementation as well as variations that dispense with the computation of extremely large axioms, thereby rendering the approach applicable albeit some completeness is lost. Moreover, we evaluate the prototype on real-world datasets.

Asynchronous federated learning (AFL) is a distributed machine learning technique that allows multiple devices to collaboratively train deep learning models without sharing local data. However, AFL suffers from low efficiency due to poor client model training quality and slow server model convergence speed, which are a result of the heterogeneous nature of both data and devices. To address these issues, we propose Efficient Asynchronous Federated Learning with Prospective Momentum Aggregation and Fine-Grained Correction (FedAC). Our framework consists of three key components. The first component is client weight evaluation based on temporal gradient, which evaluates the client weight based on the similarity between the client and server update directions. The second component is adaptive server update with prospective weighted momentum, which uses an asynchronous buffered update strategy and a prospective weighted momentum with adaptive learning rate to update the global model in server. The last component is client update with fine-grained gradient correction, which introduces a fine-grained gradient correction term to mitigate the client drift and correct the client stochastic gradient. We conduct experiments on real and synthetic datasets, and compare with existing federated learning methods. Experimental results demonstrate effective improvements in model training efficiency and AFL performance by our framework.

In this work, we focus on high-dimensional single index models with non-Gaussian sensing vectors and generative priors. More specifically, our goal is to estimate the underlying signal from i.i.d. realizations of the semi-parameterized single index model, where the underlying signal is contained in (up to a constant scaling) the range of a Lipschitz continuous generative model with bounded low-dimensional inputs, the sensing vector follows a non-Gaussian distribution, the noise is a random variable that is independent of the sensing vector, and the unknown non-linear link function is differentiable. Using the first- and second-order Stein's identity, we introduce efficient algorithms to obtain estimated vectors that achieve the near-optimal statistical rate. Experimental results on image datasets are provided to support our theory.