Machine Learning (cs.LG)

Abstract: Online learning methods yield sequential regret bounds under minimal assumptions and provide in-expectation risk bounds for statistical learning. However, despite the apparent advantage of online guarantees over their statistical counterparts, recent findings indicate that in many important cases, regret bounds may not guarantee tight high-probability risk bounds in the statistical setting. In this work we show that online to batch conversions applied to general online learning algorithms can bypass this limitation. Via a general second-order correction to the loss function defining the regret, we obtain nearly optimal high-probability risk bounds for several classical statistical estimation problems, such as discrete distribution estimation, linear regression, logistic regression, and conditional density estimation. Our analysis relies on the fact that many online learning algorithms are improper, as they are not restricted to use predictors from a given reference class. The improper nature of our estimators enables significant improvements in the dependencies on various problem parameters. Finally, we discuss some computational advantages of our sequential algorithms over their existing batch counterparts.

Abstract: Reinforcement learning has been widely successful in producing agents capable of playing games at a human level. However, this requires complex reward engineering, and the agent's resulting policy is often unpredictable. Going beyond reinforcement learning is necessary to model a wide range of human playstyles, which can be difficult to represent with a reward function. This paper presents a novel imitation learning approach to generate multiple persona policies for playtesting. Multimodal Generative Adversarial Imitation Learning (MultiGAIL) uses an auxiliary input parameter to learn distinct personas using a single-agent model. MultiGAIL is based on generative adversarial imitation learning and uses multiple discriminators as reward models, inferring the environment reward by comparing the agent and distinct expert policies. The reward from each discriminator is weighted according to the auxiliary input. Our experimental analysis demonstrates the effectiveness of our technique in two environments with continuous and discrete action spaces.

Abstract: In Chinese building codes, it is required that residential buildings receive a minimum number of hours of natural, direct sunlight on a specified winter day, which represents the worst sunlight condition in a year. This requirement is a prerequisite for obtaining a building permit during the conceptual design of a residential project. Thus, officially sanctioned software is usually used to assess the sunlight performance of buildings. These software programs predict sunlight hours based on repeated shading calculations, which is time-consuming. This paper proposed a multilayer perceptron-based method, a one-stage prediction approach, which outputs a shading time interval caused by the inputted cuboid-form building. The sunlight hours of a site can be obtained by calculating the union of the sunlight time intervals (complement of shading time interval) of all the buildings. Three numerical experiments, i.e., horizontal level and slope analysis, and simulation-based optimization are carried out; the results show that the method reduces the computation time to 1/84~1/50 with 96.5%~98% accuracies. A residential neighborhood layout planning plug-in for Rhino 7/Grasshopper is also developed based on the proposed model. This paper indicates that deep learning techniques can be adopted to accelerate sunlight hour simulations at the conceptual design phase.

Abstract: We present a simple yet novel parameterized form of linear mapping to achieves remarkable network compression performance: a pseudo SVD called Ternary SVD (TSVD). Unlike vanilla SVD, TSVD limits the $U$ and $V$ matrices in SVD to ternary matrices form in $\{\pm 1, 0\}$. This means that instead of using the expensive multiplication instructions, TSVD only requires addition instructions when computing $U(\cdot)$ and $V(\cdot)$. We provide direct and training transition algorithms for TSVD like Post Training Quantization and Quantization Aware Training respectively. Additionally, we analyze the convergence of the direct transition algorithms in theory. In experiments, we demonstrate that TSVD can achieve state-of-the-art network compression performance in various types of networks and tasks, including current baseline models such as ConvNext, Swim, BERT, and large language model like OPT.

Abstract: In recent years, the popular Transformer architecture has achieved great success in many application areas, including natural language processing and computer vision. Many existing works aim to reduce the computational and memory complexity of the self-attention mechanism in the Transformer by trading off performance. However, performance is key for the continuing success of the Transformer. In this paper, a drop-in replacement for the self-attention mechanism in the Transformer, called the Extractor, is proposed. Experimental results show that replacing the self-attention mechanism with the Extractor improves the performance of the Transformer. Furthermore, the proposed Extractor has the potential to run faster than the self-attention since it has a much shorter critical path of computation. Additionally, the sequence prediction problem in the context of text generation is formulated using variable-length discrete-time Markov chains, and the Transformer is reviewed based on our understanding.

Abstract: Quantization has become a crucial step for the efficient deployment of deep neural networks, where floating point operations are converted to simpler fixed point operations. In its most naive form, it simply consists in a combination of scaling and rounding transformations, leading to either a limited compression rate or a significant accuracy drop. Recently, Gradient-based post-training quantization (GPTQ) methods appears to be constitute a suitable trade-off between such simple methods and more powerful, yet expensive Quantization-Aware Training (QAT) approaches, particularly when attempting to quantize LLMs, where scalability of the quantization process is of paramount importance. GPTQ essentially consists in learning the rounding operation using a small calibration set. In this work, we challenge common choices in GPTQ methods. In particular, we show that the process is, to a certain extent, robust to a number of variables (weight selection, feature augmentation, choice of calibration set). More importantly, we derive a number of best practices for designing more efficient and scalable GPTQ methods, regarding the problem formulation (loss, degrees of freedom, use of non-uniform quantization schemes) or optimization process (choice of variable and optimizer). Lastly, we propose a novel importance-based mixed-precision technique. Those guidelines lead to significant performance improvements on all the tested state-of-the-art GPTQ methods and networks (e.g. +6.819 points on ViT for 4-bit quantization), paving the way for the design of scalable, yet effective quantization methods.

Abstract: Machine unlearning, the ability for a machine learning model to forget, is becoming increasingly important to comply with data privacy regulations, as well as to remove harmful, manipulated, or outdated information. The key challenge lies in forgetting specific information while protecting model performance on the remaining data. While current state-of-the-art methods perform well, they typically require some level of retraining over the retained data, in order to protect or restore model performance. This adds computational overhead and mandates that the training data remain available and accessible, which may not be feasible. In contrast, other methods employ a retrain-free paradigm, however, these approaches are prohibitively computationally expensive and do not perform on par with their retrain-based counterparts. We present Selective Synaptic Dampening (SSD), a novel two-step, post hoc, retrain-free approach to machine unlearning which is fast, performant, and does not require long-term storage of the training data. First, SSD uses the Fisher information matrix of the training and forgetting data to select parameters that are disproportionately important to the forget set. Second, SSD induces forgetting by dampening these parameters proportional to their relative importance to the forget set with respect to the wider training data. We evaluate our method against several existing unlearning methods in a range of experiments using ResNet18 and Vision Transformer. Results show that the performance of SSD is competitive with retrain-based post hoc methods, demonstrating the viability of retrain-free post hoc unlearning approaches.

Abstract: Fine-tuning pre-trained neural network models has become a widely adopted approach across various domains. However, it can lead to the distortion of pre-trained feature extractors that already possess strong generalization capabilities. Mitigating feature distortion during adaptation to new target domains is crucial. Recent studies have shown promising results in handling feature distortion by aligning the head layer on in-distribution datasets before performing fine-tuning. Nonetheless, a significant limitation arises from the treatment of batch normalization layers during fine-tuning, leading to suboptimal performance. In this paper, we propose Domain-Aware Fine-Tuning (DAFT), a novel approach that incorporates batch normalization conversion and the integration of linear probing and fine-tuning. Our batch normalization conversion method effectively mitigates feature distortion by reducing modifications to the neural network during fine-tuning. Additionally, we introduce the integration of linear probing and fine-tuning to optimize the head layer with gradual adaptation of the feature extractor. By leveraging batch normalization layers and integrating linear probing and fine-tuning, our DAFT significantly mitigates feature distortion and achieves improved model performance on both in-distribution and out-of-distribution datasets. Extensive experiments demonstrate that our method outperforms other baseline methods, demonstrating its effectiveness in not only improving performance but also mitigating feature distortion.

Abstract: Federated learning (FL) is a promising approach in distributed learning keeping privacy. However, during the training pipeline of FL, slow or incapable clients (i.e., stragglers) slow down the total training time and degrade performance. System heterogeneity, including heterogeneous computing and network bandwidth, has been addressed to mitigate the impact of stragglers. Previous studies split models to tackle the issue, but with less degree-of-freedom in terms of model architecture. We propose nested federated learning (NeFL), a generalized framework that efficiently divides a model into submodels using both depthwise and widthwise scaling. NeFL is implemented by interpreting models as solving ordinary differential equations (ODEs) with adaptive step sizes. To address the inconsistency that arises when training multiple submodels with different architecture, we decouple a few parameters. NeFL enables resource-constrained clients to effectively join the FL pipeline and the model to be trained with a larger amount of data. Through a series of experiments, we demonstrate that NeFL leads to significant gains, especially for the worst-case submodel (e.g., 8.33 improvement on CIFAR-10). Furthermore, we demonstrate NeFL aligns with recent studies in FL.

Abstract: This paper is to introduce an asynchronous and local learning framework for neural networks, named Modular Learning Framework (MOLE). This framework modularizes neural networks by layers, defines the training objective via mutual information for each module, and sequentially trains each module by mutual information maximization. MOLE makes the training become local optimization with gradient-isolated across modules, and this scheme is more biologically plausible than BP. We run experiments on vector-, grid- and graph-type data. In particular, this framework is capable of solving both graph- and node-level tasks for graph-type data. Therefore, MOLE has been experimentally proven to be universally applicable to different types of data.

Abstract: A key component of many graph neural networks (GNNs) is the pooling operation, which seeks to reduce the size of a graph while preserving important structural information. However, most existing graph pooling strategies rely on an assignment matrix obtained by employing a GNN layer, which is characterized by trainable parameters, often leading to significant computational complexity and a lack of interpretability in the pooling process. In this paper, we propose an unsupervised graph encoder-decoder model to detect abnormal nodes from graphs by learning an anomaly scoring function to rank nodes based on their degree of abnormality. In the encoding stage, we design a novel pooling mechanism, named LCPool, which leverages locality-constrained linear coding for feature encoding to find a cluster assignment matrix by solving a least-squares optimization problem with a locality regularization term. By enforcing locality constraints during the coding process, LCPool is designed to be free from learnable parameters, capable of efficiently handling large graphs, and can effectively generate a coarser graph representation while retaining the most significant structural characteristics of the graph. In the decoding stage, we propose an unpooling operation, called LCUnpool, to reconstruct both the structure and nodal features of the original graph. We conduct empirical evaluations of our method on six benchmark datasets using several evaluation metrics, and the results demonstrate its superiority over state-of-the-art anomaly detection approaches.

Abstract: Queueing systems are widely applicable stochastic models with use cases in communication networks, healthcare, service systems, etc. Although their optimal control has been extensively studied, most existing approaches assume perfect knowledge of system parameters. Of course, this assumption rarely holds in practice where there is parameter uncertainty, thus motivating a recent line of work on bandit learning for queueing systems. This nascent stream of research focuses on the asymptotic performance of the proposed algorithms. In this paper, we argue that an asymptotic metric, which focuses on late-stage performance, is insufficient to capture the intrinsic statistical complexity of learning in queueing systems which typically occurs in the early stage. Instead, we propose the Cost of Learning in Queueing (CLQ), a new metric that quantifies the maximum increase in time-averaged queue length caused by parameter uncertainty. We characterize the CLQ of a single-queue multi-server system, and then extend these results to multi-queue multi-server systems and networks of queues. In establishing our results, we propose a unified analysis framework for CLQ that bridges Lyapunov and bandit analysis, which could be of independent interest.

Abstract: The transformation towards renewable energy and feedstock supply in the chemical industry requires new conceptual process design approaches. Recently, breakthroughs in artificial intelligence offer opportunities to accelerate this transition. Specifically, deep reinforcement learning, a subclass of machine learning, has shown the potential to solve complex decision-making problems and aid sustainable process design. We survey state-of-the-art research in reinforcement learning for process design through three major elements: (i) information representation, (ii) agent architecture, and (iii) environment and reward. Moreover, we discuss perspectives on underlying challenges and promising future works to unfold the full potential of reinforcement learning for process design in chemical engineering.

Abstract: The degradation process of lithium-ion batteries is intricately linked to their entire lifecycle as power sources and energy storage devices, encompassing aspects such as performance delivery and cycling utilization. Consequently, the accurate and expedient estimation or prediction of the aging state of lithium-ion batteries has garnered extensive attention. Nonetheless, prevailing research predominantly concentrates on either aging estimation or prediction, neglecting the dynamic fusion of both facets. This paper proposes a hybrid model for capacity aging estimation and prediction based on deep learning, wherein salient features highly pertinent to aging are extracted from charge and discharge relaxation processes. By amalgamating historical capacity decay data, the model dynamically furnishes estimations of the present capacity and forecasts of future capacity for lithium-ion batteries. Our approach is validated against a novel dataset involving charge and discharge cycles at varying rates. Specifically, under a charging condition of 0.25C, a mean absolute percentage error (MAPE) of 0.29% is achieved. This outcome underscores the model's adeptness in harnessing relaxation processes commonly encountered in the real world and synergizing with historical capacity records within battery management systems (BMS), thereby affording estimations and prognostications of capacity decline with heightened precision.

Abstract: Machine learning (ML) methods are proliferating in scientific research. However, the adoption of these methods has been accompanied by failures of validity, reproducibility, and generalizability. These failures can hinder scientific progress, lead to false consensus around invalid claims, and undermine the credibility of ML-based science. ML methods are often applied and fail in similar ways across disciplines. Motivated by this observation, our goal is to provide clear reporting standards for ML-based science. Drawing from an extensive review of past literature, we present the REFORMS checklist ($\textbf{Re}$porting Standards $\textbf{For}$ $\textbf{M}$achine Learning Based $\textbf{S}$cience). It consists of 32 questions and a paired set of guidelines. REFORMS was developed based on a consensus of 19 researchers across computer science, data science, mathematics, social sciences, and biomedical sciences. REFORMS can serve as a resource for researchers when designing and implementing a study, for referees when reviewing papers, and for journals when enforcing standards for transparency and reproducibility.

Abstract: As the study of graph neural networks becomes more intensive and comprehensive, their robustness and security have received great research interest. The existing global attack methods treat all nodes in the graph as their attack targets. Although existing methods have achieved excellent results, there is still considerable space for improvement. The key problem is that the current approaches rigidly follow the definition of global attacks. They ignore an important issue, i.e., different nodes have different robustness and are not equally resilient to attacks. From a global attacker's view, we should arrange the attack budget wisely, rather than wasting them on highly robust nodes. To this end, we propose a totally new method named partial graph attack (PGA), which selects the vulnerable nodes as attack targets. First, to select the vulnerable items, we propose a hierarchical target selection policy, which allows attackers to only focus on easy-to-attack nodes. Then, we propose a cost-effective anchor-picking policy to pick the most promising anchors for adding or removing edges, and a more aggressive iterative greedy-based attack method to perform more efficient attacks. Extensive experimental results demonstrate that PGA can achieve significant improvements in both attack effect and attack efficiency compared to other existing graph global attack methods.

Abstract: Mobile health aims to enhance health outcomes by delivering interventions to individuals as they go about their daily life. The involvement of care partners and social support networks often proves crucial in helping individuals managing burdensome medical conditions. This presents opportunities in mobile health to design interventions that target the dyadic relationship -- the relationship between a target person and their care partner -- with the aim of enhancing social support. In this paper, we develop dyadic RL, an online reinforcement learning algorithm designed to personalize intervention delivery based on contextual factors and past responses of a target person and their care partner. Here, multiple sets of interventions impact the dyad across multiple time intervals. The developed dyadic RL is Bayesian and hierarchical. We formally introduce the problem setup, develop dyadic RL and establish a regret bound. We demonstrate dyadic RL's empirical performance through simulation studies on both toy scenarios and on a realistic test bed constructed from data collected in a mobile health study.

Abstract: Human emotion is expressed in many communication modalities and media formats and so their computational study is equally diversified into natural language processing, audio signal analysis, computer vision, etc. Similarly, the large variety of representation formats used in previous research to describe emotions (polarity scales, basic emotion categories, dimensional approaches, appraisal theory, etc.) have led to an ever proliferating diversity of datasets, predictive models, and software tools for emotion analysis. Because of these two distinct types of heterogeneity, at the expressional and representational level, there is a dire need to unify previous work on increasingly diverging data and label types. This article presents such a unifying computational model. We propose a training procedure that learns a shared latent representation for emotions, so-called emotion embeddings, independent of different natural languages, communication modalities, media or representation label formats, and even disparate model architectures. Experiments on a wide range of heterogeneous affective datasets indicate that this approach yields the desired interoperability for the sake of reusability, interpretability and flexibility, without penalizing prediction quality. Code and data are archived under https://doi.org/10.5281/zenodo.7405327 .

Abstract: Recently, Graph Neural Networks (GNNs) have shown promising performance in tasks on dynamic graphs such as node classification, link prediction and graph regression. However, few work has studied the temporal edge regression task which has important real-world applications. In this paper, we explore the application of GNNs to edge regression tasks in both static and dynamic settings, focusing on predicting food and agriculture trade values between nations. We introduce three simple yet strong baselines and comprehensively evaluate one static and three dynamic GNN models using the UN Trade dataset. Our experimental results reveal that the baselines exhibit remarkably strong performance across various settings, highlighting the inadequacy of existing GNNs. We also find that TGN outperforms other GNN models, suggesting TGN is a more appropriate choice for edge regression tasks. Moreover, we note that the proportion of negative edges in the training samples significantly affects the test performance. The companion source code can be found at: https://github.com/scylj1/GNN_Edge_Regression.

Abstract: The state-of-the-art in time series classification has come a long way, from the 1NN-DTW algorithm to the ROCKET family of classifiers. However, in the current fast-paced development of new classifiers, taking a step back and performing simple baseline checks is essential. These checks are often overlooked, as researchers are focused on establishing new state-of-the-art results, developing scalable algorithms, and making models explainable. Nevertheless, there are many datasets that look like time series at first glance, but classic algorithms such as tabular methods with no time ordering may perform better on such problems. For example, for spectroscopy datasets, tabular methods tend to significantly outperform recent time series methods. In this study, we compare the performance of tabular models using classic machine learning approaches (e.g., Ridge, LDA, RandomForest) with the ROCKET family of classifiers (e.g., Rocket, MiniRocket, MultiRocket). Tabular models are simple and very efficient, while the ROCKET family of classifiers are more complex and have state-of-the-art accuracy and efficiency among recent time series classifiers. We find that tabular models outperform the ROCKET family of classifiers on approximately 19% of univariate and 28% of multivariate datasets in the UCR/UEA benchmark and achieve accuracy within 10 percentage points on about 50% of datasets. Our results suggest that it is important to consider simple tabular models as baselines when developing time series classifiers. These models are very fast, can be as effective as more complex methods and may be easier to understand and deploy.

Abstract: We propose \emph{regular expression inference (REI)} as a challenge for code/language modelling, and the wider machine learning community. REI is a supervised machine learning (ML) and program synthesis task, and poses the problem of finding minimal regular expressions from examples: Given two finite sets of strings $P$ and $N$ and a cost function $\text{cost}(\cdot)$, the task is to generate an expression $r$ that accepts all strings in $P$ and rejects all strings in $N$, while no other such expression $r'$ exists with $\text{cost}(r')<\text{cost}(r)$. REI has advantages as a challenge problem: (i) regular expressions are well-known, widely used, and a natural idealisation of code; (ii) REI's asymptotic worst-case complexity is well understood; (iii) REI has a small number of easy to understand parameters (e.g.~$P$ or $N$ cardinality, string lengths of examples, or the cost function); this lets us easily finetune REI-hardness; (iv) REI is an unsolved problem for deep learning based ML. Recently, an REI solver was implemented on GPUs, using program synthesis techniques. This enabled, for the first time, fast generation of minimal expressions for complex REI instances. Building on this advance, we generate and publish the first large-scale datasets for REI, and devise and evaluate several initial heuristic and machine learning baselines. We invite the community to participate and explore ML methods that learn to solve REI problems. We believe that progress in REI directly translates to code/language modelling.