Machine Learning (cs.LG)

Abstract: We prove the converse of the universal approximation theorem, i.e. a neural network (NN) encoding theorem which shows that for every stably converged NN of continuous activation functions, its weight matrix actually encodes a continuous function that approximates its training dataset to within a finite margin of error over a bounded domain. We further show that using the Eckart-Young theorem for truncated singular value decomposition of the weight matrix for every NN layer, we can illuminate the nature of the latent space manifold of the training dataset encoded and represented by every NN layer, and the geometric nature of the mathematical operations performed by each NN layer. Our results have implications for understanding how NNs break the curse of dimensionality by harnessing memory capacity for expressivity, and that the two are complementary. This Layer Matrix Decomposition (LMD) further suggests a close relationship between eigen-decomposition of NN layers and the latest advances in conceptualizations of Hopfield networks and Transformer NN models.

Abstract: In this work, we present CleanUNet 2, a speech denoising model that combines the advantages of waveform denoiser and spectrogram denoiser and achieves the best of both worlds. CleanUNet 2 uses a two-stage framework inspired by popular speech synthesis methods that consist of a waveform model and a spectrogram model. Specifically, CleanUNet 2 builds upon CleanUNet, the state-of-the-art waveform denoiser, and further boosts its performance by taking predicted spectrograms from a spectrogram denoiser as the input. We demonstrate that CleanUNet 2 outperforms previous methods in terms of various objective and subjective evaluations.

Abstract: Quantification of the political leaning of online news articles can aid in understanding the dynamics of political ideology in social groups and measures to mitigating them. However, predicting the accurate political leaning of a news article with machine learning models is a challenging task. This is due to (i) the political ideology of a news article is defined by several factors, and (ii) the innate nature of existing learning models to be biased with the political bias of the news publisher during the model training. There is only a limited number of methods to study the political leaning of news articles which also do not consider the algorithmic political bias which lowers the generalization of machine learning models to predict the political leaning of news articles published by any new news publishers. In this work, we propose a knowledge-infused deep learning model that utilizes relatively reliable external data resources to learn unbiased representations of news articles using their global and local contexts. We evaluate the proposed model by setting the data in such a way that news domains or news publishers in the test set are completely unseen during the training phase. With this setup we show that the proposed model mitigates algorithmic political bias and outperforms baseline methods to predict the political leaning of news articles with up to 73% accuracy.

Abstract: We investigate the expressive power of deep residual neural networks idealized as continuous dynamical systems through control theory. Specifically, we consider two properties that arise from supervised learning, namely universal interpolation - the ability to match arbitrary input and target training samples - and the closely related notion of universal approximation - the ability to approximate input-target functional relationships via flow maps. Under the assumption of affine invariance of the control family, we give a characterisation of universal interpolation, showing that it holds for essentially any architecture with non-linearity. Furthermore, we elucidate the relationship between universal interpolation and universal approximation in the context of general control systems, showing that the two properties cannot be deduced from each other. At the same time, we identify conditions on the control family and the target function that ensures the equivalence of the two notions.

Abstract: In different wireless network scenarios, multiple network entities need to cooperate in order to achieve a common task with minimum delay and energy consumption. Future wireless networks mandate exchanging high dimensional data in dynamic and uncertain environments, therefore implementing communication control tasks becomes challenging and highly complex. Multi-agent reinforcement learning with emergent communication (EC-MARL) is a promising solution to address high dimensional continuous control problems with partially observable states in a cooperative fashion where agents build an emergent communication protocol to solve complex tasks. This paper articulates the importance of EC-MARL within the context of future 6G wireless networks, which imbues autonomous decision-making capabilities into network entities to solve complex tasks such as autonomous driving, robot navigation, flying base stations network planning, and smart city applications. An overview of EC-MARL algorithms and their design criteria are provided while presenting use cases and research opportunities on this emerging topic.

Abstract: Graph anomaly detection (GAD) has attracted increasing attention in machine learning and data mining. Recent works have mainly focused on how to capture richer information to improve the quality of node embeddings for GAD. Despite their significant advances in detection performance, there is still a relative dearth of research on the properties of the task. GAD aims to discern the anomalies that deviate from most nodes. However, the model is prone to learn the pattern of normal samples which make up the majority of samples. Meanwhile, anomalies can be easily detected when their behaviors differ from normality. Therefore, the performance can be further improved by enhancing the ability to learn the normal pattern. To this end, we propose a normality learning-based GAD framework via multi-scale contrastive learning networks (NLGAD for abbreviation). Specifically, we first initialize the model with the contrastive networks on different scales. To provide sufficient and reliable normal nodes for normality learning, we design an effective hybrid strategy for normality selection. Finally, the model is refined with the only input of reliable normal nodes and learns a more accurate estimate of normality so that anomalous nodes can be more easily distinguished. Eventually, extensive experiments on six benchmark graph datasets demonstrate the effectiveness of our normality learning-based scheme on GAD. Notably, the proposed algorithm improves the detection performance (up to 5.89% AUC gain) compared with the state-of-the-art methods. The source code is released at https://github.com/FelixDJC/NLGAD.

Abstract: The negative impact of label noise is well studied in classical supervised learning yet remains an open research question in meta-learning. Meta-learners aim to adapt to unseen learning tasks by learning a good initial model in meta-training and consecutively fine-tuning it according to new tasks during meta-testing. In this paper, we present the first extensive analysis of the impact of varying levels of label noise on the performance of state-of-the-art meta-learners, specifically gradient-based $N$-way $K$-shot learners. We show that the accuracy of Reptile, iMAML, and foMAML drops by up to 42% on the Omniglot and CifarFS datasets when meta-training is affected by label noise. To strengthen the resilience against label noise, we propose two sampling techniques, namely manifold (Man) and batch manifold (BatMan), which transform the noisy supervised learners into semi-supervised ones to increase the utility of noisy labels. We first construct manifold samples of $N$-way $2$-contrastive-shot tasks through augmentation, learning the embedding via a contrastive loss in meta-training, and then perform classification through zeroing on the embedding in meta-testing. We show that our approach can effectively mitigate the impact of meta-training label noise. Even with 60% wrong labels \batman and \man can limit the meta-testing accuracy drop to ${2.5}$, ${9.4}$, ${1.1}$ percent points, respectively, with existing meta-learners across the Omniglot, CifarFS, and MiniImagenet datasets.

Abstract: This paper presents a novel online learning method that aims at finding a separator hyperplane between data points labelled as either positive or negative. Since weights and biases of artificial neurons can directly be related to hyperplanes in high-dimensional spaces, the technique is applicable to train perceptron-based binary classifiers in machine learning. In case of large or imbalanced data sets, use of analytical or gradient-based solutions can become prohibitive and impractical, where heuristics and approximation techniques are still applicable. The proposed method is based on the Perceptron algorithm, however, it tunes neuron weights in just the necessary extent during searching the separator hyperplane. Due to an appropriate transformation of the initial data set we need not to consider data labels, neither the bias term. respectively, reducing separability to a one-class classification problem. The presented method has proven converge; empirical results show that it can be more efficient than the Perceptron algorithm, especially, when the size of the data set exceeds data dimensionality.

Abstract: In-context learning, i.e., learning from in-context samples, is an impressive ability of Transformer. However, the mechanism driving the in-context learning is not yet fully understood. In this study, we aim to investigate from an underexplored perspective of representation learning. The representation is more complex for in-context learning senario, where the representation can be impacted by both model weights and in-context samples. We refer the above two conceptually aspects of representation as in-weight component and in-context component, respectively. To study how the two components affect in-context learning capabilities, we construct a novel synthetic task, making it possible to device two probes, in-weights probe and in-context probe, to evaluate the two components, respectively. We demonstrate that the goodness of in-context component is highly related to the in-context learning performance, which indicates the entanglement between in-context learning and representation learning. Furthermore, we find that a good in-weights component can actually benefit the learning of the in-context component, indicating that in-weights learning should be the foundation of in-context learning. To further understand the the in-context learning mechanism and importance of the in-weights component, we proof by construction that a simple Transformer, which uses pattern matching and copy-past mechanism to perform in-context learning, can match the in-context learning performance with more complex, best tuned Transformer under the perfect in-weights component assumption. In short, those discoveries from representation learning perspective shed light on new approaches to improve the in-context capacity.

Abstract: Landslides are a common natural disaster that can cause casualties, property safety threats and economic losses. Therefore, it is important to understand or predict the probability of landslide occurrence at potentially risky sites. A commonly used means is to carry out a landslide susceptibility assessment based on a landslide inventory and a set of landslide contributing factors. This can be readily achieved using machine learning (ML) models such as logistic regression (LR), support vector machine (SVM), random forest (RF), extreme gradient boosting (Xgboost), or deep learning (DL) models such as convolutional neural network (CNN) and long short time memory (LSTM). As the input data for these models, landslide contributing factors have varying influences on landslide occurrence. Therefore, it is logically feasible to select more important contributing factors and eliminate less relevant ones, with the aim of increasing the prediction accuracy of these models. However, selecting more important factors is still a challenging task and there is no generally accepted method. Furthermore, the effects of factor selection using various methods on the prediction accuracy of ML and DL models are unclear. In this study, the impact of the selection of contributing factors on the accuracy of landslide susceptibility predictions using ML and DL models was investigated. Four methods for selecting contributing factors were considered for all the aforementioned ML and DL models, which included Information Gain Ratio (IGR), Recursive Feature Elimination (RFE), Particle Swarm Optimization (PSO), Least Absolute Shrinkage and Selection Operators (LASSO) and Harris Hawk Optimization (HHO). In addition, autoencoder-based factor selection methods for DL models were also investigated. To assess their performances, an exhaustive approach was adopted,...

Abstract: Graph Neural Networks (GNNs) have gained popularity in healthcare and other domains due to their ability to process multi-modal and multi-relational graphs. However, efficient training of GNNs remains challenging, with several open research questions. In this paper, we investigate how the flow of embedding information within GNNs affects the prediction of links in Knowledge Graphs (KGs). Specifically, we propose a mathematical model that decouples the GNN connectivity from the connectivity of the graph data and evaluate the performance of GNNs in a clinical triage use case. Our results demonstrate that incorporating domain knowledge into the GNN connectivity leads to better performance than using the same connectivity as the KG or allowing unconstrained embedding propagation. Moreover, we show that negative edges play a crucial role in achieving good predictions, and that using too many GNN layers can degrade performance.

Abstract: Power grids are moving towards 100% renewable energy source bulk power grids, and the overall dynamics of power system operations and electricity markets are changing. The electricity markets are not only dispatching resources economically but also taking into account various controllable actions like renewable curtailment, transmission congestion mitigation, and energy storage optimization to ensure grid reliability. As a result, price formations in electricity markets have become quite complex. Traditional root cause analysis and statistical approaches are rendered inapplicable to analyze and infer the main drivers behind price formation in the modern grid and markets with variable renewable energy (VRE). In this paper, we propose a machine learning-based analysis framework to deconstruct the primary drivers for price spike events in modern electricity markets with high renewable energy. The outcomes can be utilized for various critical aspects of market design, renewable dispatch and curtailment, operations, and cyber-security applications. The framework can be applied to any ISO or market data; however, in this paper, it is applied to open-source publicly available datasets from California Independent System Operator (CAISO) and ISO New England (ISO-NE).

Abstract: Continuous unsupervised representation learning (CURL) research has greatly benefited from improvements in self-supervised learning (SSL) techniques. As a result, existing CURL methods using SSL can learn high-quality representations without any labels, but with a notable performance drop when learning on a many-tasks data stream. We hypothesize that this is caused by the regularization losses that are imposed to prevent forgetting, leading to a suboptimal plasticity-stability trade-off: they either do not adapt fully to the incoming data (low plasticity), or incur significant forgetting when allowed to fully adapt to a new SSL pretext-task (low stability). In this work, we propose to train an expert network that is relieved of the duty of keeping the previous knowledge and can focus on performing optimally on the new tasks (optimizing plasticity). In the second phase, we combine this new knowledge with the previous network in an adaptation-retrospection phase to avoid forgetting and initialize a new expert with the knowledge of the old network. We perform several experiments showing that our proposed approach outperforms other CURL exemplar-free methods in few- and many-task split settings. Furthermore, we show how to adapt our approach to semi-supervised continual learning (Semi-SCL) and show that we surpass the accuracy of other exemplar-free Semi-SCL methods and reach the results of some others that use exemplars.

Abstract: In this paper, a Robust Multi-branch Deep learning-based system for remaining useful life (RUL) prediction and condition operations (CO) identification of rotating machines is proposed. In particular, the proposed system comprises main components: (1) an LSTM-Autoencoder to denoise the vibration data; (2) a feature extraction to generate time-domain, frequency-domain, and time-frequency based features from the denoised data; (3) a novel and robust multi-branch deep learning network architecture to exploit the multiple features. The performance of our proposed system was evaluated and compared to the state-of-the-art systems on two benchmark datasets of XJTU-SY and PRONOSTIA. The experimental results prove that our proposed system outperforms the state-of-the-art systems and presents potential for real-life applications on bearing machines.

Abstract: Despite recent success, state-of-the-art learning-based models remain highly vulnerable to input changes such as adversarial examples. In order to obtain certifiable robustness against such perturbations, recent work considers Lipschitz-based regularizers or constraints while at the same time increasing prediction margin. Unfortunately, this comes at the cost of significantly decreased accuracy. In this paper, we propose a Calibrated Lipschitz-Margin Loss (CLL) that addresses this issue and improves certified robustness by tackling two problems: Firstly, commonly used margin losses do not adjust the penalties to the shrinking output distribution; caused by minimizing the Lipschitz constant $K$. Secondly, and most importantly, we observe that minimization of $K$ can lead to overly smooth decision functions. This limits the model's complexity and thus reduces accuracy. Our CLL addresses these issues by explicitly calibrating the loss w.r.t. margin and Lipschitz constant, thereby establishing full control over slack and improving robustness certificates even with larger Lipschitz constants. On CIFAR-10, CIFAR-100 and Tiny-ImageNet, our models consistently outperform losses that leave the constant unattended. On CIFAR-100 and Tiny-ImageNet, CLL improves upon state-of-the-art deterministic $L_2$ robust accuracies. In contrast to current trends, we unlock potential of much smaller models without $K=1$ constraints.

Abstract: Diffusion models (DMs) demonstrate potent image generation capabilities in various generative modeling tasks. Nevertheless, their primary limitation lies in slow sampling speed, requiring hundreds or thousands of sequential function evaluations through large neural networks to generate high-quality images. Sampling from DMs can be seen as solving corresponding stochastic differential equations (SDEs) or ordinary differential equations (ODEs). In this work, we formulate the sampling process as an extended reverse-time SDE (ER SDE), unifying prior explorations into ODEs and SDEs. Leveraging the semi-linear structure of ER SDE solutions, we offer exact solutions and arbitrarily high-order approximate solutions for VP SDE and VE SDE, respectively. Based on the solution space of the ER SDE, we yield mathematical insights elucidating the superior performance of ODE solvers over SDE solvers in terms of fast sampling. Additionally, we unveil that VP SDE solvers stand on par with their VE SDE counterparts. Finally, we devise fast and training-free samplers, ER-SDE Solvers, elevating the efficiency of stochastic samplers to unprecedented levels. Experimental results demonstrate achieving 3.45 FID in 20 function evaluations and 2.24 FID in 50 function evaluations on the ImageNet 64$\times$64 dataset.

Abstract: High throughput serving of large language models (LLMs) requires batching sufficiently many requests at a time. However, existing systems struggle because the key-value cache (KV cache) memory for each request is huge and grows and shrinks dynamically. When managed inefficiently, this memory can be significantly wasted by fragmentation and redundant duplication, limiting the batch size. To address this problem, we propose PagedAttention, an attention algorithm inspired by the classical virtual memory and paging techniques in operating systems. On top of it, we build vLLM, an LLM serving system that achieves (1) near-zero waste in KV cache memory and (2) flexible sharing of KV cache within and across requests to further reduce memory usage. Our evaluations show that vLLM improves the throughput of popular LLMs by 2-4$\times$ with the same level of latency compared to the state-of-the-art systems, such as FasterTransformer and Orca. The improvement is more pronounced with longer sequences, larger models, and more complex decoding algorithms. vLLM's source code is publicly available at https://github.com/vllm-project/vllm

Abstract: Solving linear inverse problems plays a crucial role in numerous applications. Algorithm unfolding based, model-aware data-driven approaches have gained significant attention for effectively addressing these problems. Learned iterative soft-thresholding algorithm (LISTA) and alternating direction method of multipliers compressive sensing network (ADMM-CSNet) are two widely used such approaches, based on ISTA and ADMM algorithms, respectively. In this work, we study optimization guarantees, i.e., achieving near-zero training loss with the increase in the number of learning epochs, for finite-layer unfolded networks such as LISTA and ADMM-CSNet with smooth soft-thresholding in an over-parameterized (OP) regime. We achieve this by leveraging a modified version of the Polyak-Lojasiewicz, denoted PL$^*$, condition. Satisfying the PL$^*$ condition within a specific region of the loss landscape ensures the existence of a global minimum and exponential convergence from initialization using gradient descent based methods. Hence, we provide conditions, in terms of the network width and the number of training samples, on these unfolded networks for the PL$^*$ condition to hold. We achieve this by deriving the Hessian spectral norm of these networks. Additionally, we show that the threshold on the number of training samples increases with the increase in the network width. Furthermore, we compare the threshold on training samples of unfolded networks with that of a standard fully-connected feed-forward network (FFNN) with smooth soft-thresholding non-linearity. We prove that unfolded networks have a higher threshold value than FFNN. Consequently, one can expect a better expected error for unfolded networks than FFNN.

Abstract: The accurate prediction of drought probability in specific regions is crucial for informed decision-making in agricultural practices. It is important to make predictions one year in advance, particularly for long-term decisions. However, forecasting this probability presents challenges due to the complex interplay of various factors within the region of interest and neighboring areas. In this study, we propose an end-to-end solution to address this issue based on various spatiotemporal neural networks. The models considered focus on predicting the drought intensity based on the Palmer Drought Severity Index (PDSI) for subregions of interest, leveraging intrinsic factors and insights from climate models to enhance drought predictions. Comparative evaluations demonstrate the superior accuracy of Convolutional LSTM (ConvLSTM) and transformer models compared to baseline gradient boosting and logistic regression solutions. The two former models achieved impressive ROC AUC scores from 0.90 to 0.70 for forecast horizons from one to six months, outperforming baseline models. The transformer showed superiority for shorter horizons, while ConvLSTM did so for longer horizons. Thus, we recommend selecting the models accordingly for long-term drought forecasting. To ensure the broad applicability of the considered models, we conduct extensive validation across regions worldwide, considering different environmental conditions. We also run several ablation and sensitivity studies to challenge our findings and provide additional information on how to solve the problem.

Abstract: Large Language Models (LLMs) have demonstrated remarkable generalization across diverse tasks, leading individuals to increasingly use them as personal assistants and universal computing engines. Nevertheless, a notable obstacle emerges when feeding numerical/temporal data into these models, such as data sourced from wearables or electronic health records. LLMs employ tokenizers in their input that break down text into smaller units. However, tokenizers are not designed to represent numerical values and might struggle to understand repetitive patterns and context, treating consecutive values as separate tokens and disregarding their temporal relationships. Here, we discuss recent works that employ LLMs for human-centric tasks such as in mobile health sensing and present a case study showing that popular LLMs tokenize temporal data incorrectly. To address that, we highlight potential solutions such as prompt tuning with lightweight embedding layers as well as multimodal adapters, that can help bridge this "modality gap". While the capability of language models to generalize to other modalities with minimal or no finetuning is exciting, this paper underscores the fact that their outputs cannot be meaningful if they stumble over input nuances.

Abstract: In the dynamic and uncertain environments where reinforcement learning (RL) operates, risk management becomes a crucial factor in ensuring reliable decision-making. Traditional RL approaches, while effective in reward optimization, often overlook the landscape of potential risks. In response, this paper pioneers the integration of Optimal Transport (OT) theory with RL to create a risk-aware framework. Our approach modifies the objective function, ensuring that the resulting policy not only maximizes expected rewards but also respects risk constraints dictated by OT distances between state visitation distributions and the desired risk profiles. By leveraging the mathematical precision of OT, we offer a formulation that elevates risk considerations alongside conventional RL objectives. Our contributions are substantiated with a series of theorems, mapping the relationships between risk distributions, optimal value functions, and policy behaviors. Through the lens of OT, this work illuminates a promising direction for RL, ensuring a balanced fusion of reward pursuit and risk awareness.

Abstract: Probability estimation models play an important role in various fields, such as weather forecasting, recommendation systems, and sports analysis. Among several models estimating probabilities, it is difficult to evaluate which model gives reliable probabilities since the ground-truth probabilities are not available. The win probability estimation model for esports, which calculates the win probability under a certain game state, is also one of the fields being actively studied in probability estimation. However, most of the previous works evaluated their models using accuracy, a metric that only can measure the performance of discrimination. In this work, we firstly investigate the Brier score and the Expected Calibration Error (ECE) as a replacement of accuracy used as a performance evaluation metric for win probability estimation models in esports field. Based on the analysis, we propose a novel metric called Balance score which is a simple yet effective metric in terms of six good properties that probability estimation metric should have. Under the general condition, we also found that the Balance score can be an effective approximation of the true expected calibration error which has been imperfectly approximated by ECE using the binning technique. Extensive evaluations using simulation studies and real game snapshot data demonstrate the promising potential to adopt the proposed metric not only for the win probability estimation model for esports but also for evaluating general probability estimation models.

Abstract: Foundation models, including Vision Language Models (VLMs) and Large Language Models (LLMs), possess the $generality$ to handle diverse distributions and tasks, which stems from their extensive pre-training datasets. The fine-tuning of foundation models is a common practice to enhance task performance or align the model's behavior with human expectations, allowing them to gain $speciality$. However, the small datasets used for fine-tuning may not adequately cover the diverse distributions and tasks encountered during pre-training. Consequently, the pursuit of speciality during fine-tuning can lead to a loss of {generality} in the model, which is related to catastrophic forgetting (CF) in deep learning. In this study, we demonstrate this phenomenon in both VLMs and LLMs. For instance, fine-tuning VLMs like CLIP on ImageNet results in a loss of generality in handling diverse distributions, and fine-tuning LLMs like Galactica in the medical domain leads to a loss in following instructions and common sense. To address the trade-off between the speciality and generality, we investigate multiple regularization methods from continual learning, the weight averaging method (Wise-FT) from out-of-distributional (OOD) generalization, which interpolates parameters between pre-trained and fine-tuned models, and parameter-efficient fine-tuning methods like Low-Rank Adaptation (LoRA). Our findings show that both continual learning and Wise-ft methods effectively mitigate the loss of generality, with Wise-FT exhibiting the strongest performance in balancing speciality and generality.

Abstract: We present a novel, fast (exponential rate adaption), ab initio (hyper-parameter-free) gradient based optimizer algorithm. The main idea of the method is to adapt the learning rate $\alpha$ by situational awareness, mainly striving for orthogonal neighboring gradients. The method has a high success and fast convergence rate and does not rely on hand-tuned parameters giving it greater universality. It can be applied to problems of any dimensions n and scales only linearly (of order O(n)) with the dimension of the problem. It optimizes convex and non-convex continuous landscapes providing some kind of gradient. In contrast to the Ada-family (AdaGrad, AdaMax, AdaDelta, Adam, etc.) the method is rotation invariant: optimization path and performance are independent of coordinate choices. The impressive performance is demonstrated by extensive experiments on the MNIST benchmark data-set against state-of-the-art optimizers. We name this new class of optimizers after its core idea Exponential Learning Rate Adaption - ELRA. We present it in two variants c2min and p2min with slightly different control. The authors strongly believe that ELRA will open a completely new research direction for gradient descent optimize.

Abstract: The Harrisonburg Department of Public Transportation (HDPT) aims to leverage their data to improve the efficiency and effectiveness of their operations. We construct two supply and demand models that help the department identify gaps in their service. The models take many variables into account, including the way that the HDPT reports to the federal government and the areas with the most vulnerable populations in Harrisonburg City. We employ data analysis and machine learning techniques to make our predictions.

Abstract: A set of variables is the Markov blanket of a random variable if it contains all the information needed for predicting the variable. If the blanket cannot be reduced without losing useful information, it is called a Markov boundary. Identifying the Markov boundary of a random variable is advantageous because all variables outside the boundary are superfluous. Hence, the Markov boundary provides an optimal feature set. However, learning the Markov boundary from data is challenging for two reasons. If one or more variables are removed from the Markov boundary, variables outside the boundary may start providing information. Conversely, variables within the boundary may stop providing information. The true role of each candidate variable is only manifesting when the Markov boundary has been identified. In this paper, we propose a new Tsetlin Machine (TM) feedback scheme that supplements Type I and Type II feedback. The scheme introduces a novel Finite State Automaton - a Context-Specific Independence Automaton. The automaton learns which features are outside the Markov boundary of the target, allowing them to be pruned from the TM during learning. We investigate the new scheme empirically, showing how it is capable of exploiting context-specific independence to find Markov boundaries. Further, we provide a theoretical analysis of convergence. Our approach thus connects the field of Bayesian networks (BN) with TMs, potentially opening up for synergies when it comes to inference and learning, including TM-produced Bayesian knowledge bases and TM-based Bayesian inference.

Abstract: When deploying classifiers in the real world, users expect them to respond to inputs appropriately. However, traditional classifiers are not equipped to handle inputs which lie far from the distribution they were trained on. Malicious actors can exploit this defect by making adversarial perturbations designed to cause the classifier to give an incorrect output. Classification-with-rejection methods attempt to solve this problem by allowing networks to refuse to classify an input in which they have low confidence. This works well for strongly adversarial examples, but also leads to the rejection of weakly perturbed images, which intuitively could be correctly classified. To address these issues, we propose Reed-Muller Aggregation Networks (RMAggNet), a classifier inspired by Reed-Muller error-correction codes which can correct and reject inputs. This paper shows that RMAggNet can minimise incorrectness while maintaining good correctness over multiple adversarial attacks at different perturbation budgets by leveraging the ability to correct errors in the classification process. This provides an alternative classification-with-rejection method which can reduce the amount of additional processing in situations where a small number of incorrect classifications are permissible.

Abstract: Diffusion models have revolutionized text-to-image generation with its exceptional quality and creativity. However, its multi-step sampling process is known to be slow, often requiring tens of inference steps to obtain satisfactory results. Previous attempts to improve its sampling speed and reduce computational costs through distillation have been unsuccessful in achieving a functional one-step model. In this paper, we explore a recent method called Rectified Flow, which, thus far, has only been applied to small datasets. The core of Rectified Flow lies in its \emph{reflow} procedure, which straightens the trajectories of probability flows, refines the coupling between noises and images, and facilitates the distillation process with student models. We propose a novel text-conditioned pipeline to turn Stable Diffusion (SD) into an ultra-fast one-step model, in which we find reflow plays a critical role in improving the assignment between noise and images. Leveraging our new pipeline, we create, to the best of our knowledge, the first one-step diffusion-based text-to-image generator with SD-level image quality, achieving an FID (Frechet Inception Distance) of $23.3$ on MS COCO 2017-5k, surpassing the previous state-of-the-art technique, progressive distillation, by a significant margin ($37.2$ $\rightarrow$ $23.3$ in FID). By utilizing an expanded network with 1.7B parameters, we further improve the FID to $22.4$. We call our one-step models \emph{InstaFlow}. On MS COCO 2014-30k, InstaFlow yields an FID of $13.1$ in just $0.09$ second, the best in $\leq 0.1$ second regime, outperforming the recent StyleGAN-T ($13.9$ in $0.1$ second). Notably, the training of InstaFlow only costs 199 A100 GPU days. Project page:~\url{https://github.com/gnobitab/InstaFlow}.

Abstract: Can an $\mathbb{R}^n\rightarrow \mathbb{R}^n$ feedforward network learn matrix-vector multiplication? This study introduces two mechanisms - flexible masking to take matrix inputs, and a unique network pruning to respect the mask's dependency structure. Networks can approximate fixed operations such as matrix-vector multiplication $\phi(A,x) \rightarrow Ax$, motivating the mechanisms introduced with applications towards litmus-testing dependencies or interaction order in graph-based models.

Abstract: We consider the classical problem of learning, with arbitrary accuracy, the natural parameters of a $k$-parameter truncated \textit{minimal} exponential family from i.i.d. samples in a computationally and statistically efficient manner. We focus on the setting where the support as well as the natural parameters are appropriately bounded. While the traditional maximum likelihood estimator for this class of exponential family is consistent, asymptotically normal, and asymptotically efficient, evaluating it is computationally hard. In this work, we propose a novel loss function and a computationally efficient estimator that is consistent as well as asymptotically normal under mild conditions. We show that, at the population level, our method can be viewed as the maximum likelihood estimation of a re-parameterized distribution belonging to the same class of exponential family. Further, we show that our estimator can be interpreted as a solution to minimizing a particular Bregman score as well as an instance of minimizing the \textit{surrogate} likelihood. We also provide finite sample guarantees to achieve an error (in $\ell_2$-norm) of $\alpha$ in the parameter estimation with sample complexity $O({\sf poly}(k)/\alpha^2)$. Our method achives the order-optimal sample complexity of $O({\sf log}(k)/\alpha^2)$ when tailored for node-wise-sparse Markov random fields. Finally, we demonstrate the performance of our estimator via numerical experiments.