Machine Learning (cs.LG)

Abstract: Sharpness-aware minimization (SAM), which searches for flat minima by min-max optimization, has been shown to be useful in improving model generalization. However, since each SAM update requires computing two gradients, its computational cost and training time are both doubled compared to standard empirical risk minimization (ERM). Recent state-of-the-arts reduce the fraction of SAM updates and thus accelerate SAM by switching between SAM and ERM updates randomly or periodically. In this paper, we design an adaptive policy to employ SAM based on the loss landscape geometry. Two efficient algorithms, AE-SAM and AE-LookSAM, are proposed. We theoretically show that AE-SAM has the same convergence rate as SAM. Experimental results on various datasets and architectures demonstrate the efficiency and effectiveness of the adaptive policy.

Abstract: Machine and deep learning methods for medical and healthcare applications have shown significant progress and performance improvement in recent years. These methods require vast amounts of training data which are available in the medical sector, albeit decentralized. Medical institutions generate vast amounts of data for which sharing and centralizing remains a challenge as the result of data and privacy regulations. The federated learning technique is well-suited to tackle these challenges. However, federated learning comes with a new set of open problems related to communication overhead, efficient parameter aggregation, client selection strategies and more. In this work, we address the step prior to the initiation of a federated network for model training, client recruitment. By intelligently recruiting clients, communication overhead and overall cost of training can be reduced without sacrificing predictive performance. Client recruitment aims at pre-excluding potential clients from partaking in the federation based on a set of criteria indicative of their eventual contributions to the federation. In this work, we propose a client recruitment approach using only the output distribution and sample size at the client site. We show how a subset of clients can be recruited without sacrificing model performance whilst, at the same time, significantly improving computation time. By applying the recruitment approach to the training of federated models for accurate patient Length of Stay prediction using data from 189 Intensive Care Units, we show how the models trained in federations made up from recruited clients significantly outperform federated models trained with the standard procedure in terms of predictive power and training time.

Abstract: As phasor measurement units (PMUs) become more widely used in transmission power systems, a fast state estimation (SE) algorithm that can take advantage of their high sample rates is needed. To accomplish this, we present a method that uses graph neural networks (GNNs) to learn complex bus voltage estimates from PMU voltage and current measurements. We propose an original implementation of GNNs over the power system's factor graph to simplify the integration of various types and quantities of measurements on power system buses and branches. Furthermore, we augment the factor graph to improve the robustness of GNN predictions. This model is highly efficient and scalable, as its computational complexity is linear with respect to the number of nodes in the power system. Training and test examples were generated by randomly sampling sets of power system measurements and annotated with the exact solutions of linear SE with PMUs. The numerical results demonstrate that the GNN model provides an accurate approximation of the SE solutions. Furthermore, errors caused by PMU malfunctions or communication failures that would normally make the SE problem unobservable have a local effect and do not deteriorate the results in the rest of the power system.

Abstract: Tensor graph superoptimisation systems perform a sequence of subgraph substitution to neural networks, to find the optimal computation graph structure. Such a graph transformation process naturally falls into the framework of sequential decision-making, and existing systems typically employ a greedy search approach, which cannot explore the whole search space as it cannot tolerate a temporary loss of performance. In this paper, we address the tensor graph superoptimisation problem by exploring an alternative search approach, reinforcement learning (RL). Our proposed approach, X-RLflow, can learn to perform neural network dataflow graph rewriting, which substitutes a subgraph one at a time. X-RLflow is based on a model-free RL agent that uses a graph neural network (GNN) to encode the target computation graph and outputs a transformed computation graph iteratively. We show that our approach can outperform state-of-the-art superoptimisation systems over a range of deep learning models and achieve by up to 40% on those that are based on transformer-style architectures.

Abstract: Price forecasting for used construction equipment is a challenging task due to spatial and temporal price fluctuations. It is thus of high interest to automate the forecasting process based on current market data. Even though applying machine learning (ML) to these data represents a promising approach to predict the residual value of certain tools, it is hard to implement for small and medium-sized enterprises due to their insufficient ML expertise. To this end, we demonstrate the possibility of substituting manually created ML pipelines with automated machine learning (AutoML) solutions, which automatically generate the underlying pipelines. We combine AutoML methods with the domain knowledge of the companies. Based on the CRISP-DM process, we split the manual ML pipeline into a machine learning and non-machine learning part. To take all complex industrial requirements into account and to demonstrate the applicability of our new approach, we designed a novel metric named method evaluation score, which incorporates the most important technical and non-technical metrics for quality and usability. Based on this metric, we show in a case study for the industrial use case of price forecasting, that domain knowledge combined with AutoML can weaken the dependence on ML experts for innovative small and medium-sized enterprises which are interested in conducting such solutions.

Abstract: Self-training based semi-supervised learning algorithms have enabled the learning of highly accurate deep neural networks, using only a fraction of labeled data. However, the majority of work on self-training has focused on the objective of improving accuracy, whereas practical machine learning systems can have complex goals (e.g. maximizing the minimum of recall across classes, etc.) that are non-decomposable in nature. In this work, we introduce the Cost-Sensitive Self-Training (CSST) framework which generalizes the self-training-based methods for optimizing non-decomposable metrics. We prove that our framework can better optimize the desired non-decomposable metric utilizing unlabeled data, under similar data distribution assumptions made for the analysis of self-training. Using the proposed CSST framework, we obtain practical self-training methods (for both vision and NLP tasks) for optimizing different non-decomposable metrics using deep neural networks. Our results demonstrate that CSST achieves an improvement over the state-of-the-art in majority of the cases across datasets and objectives.

Abstract: Well-tuned hyperparameters are crucial for obtaining good generalization behavior in neural networks. They can enforce appropriate inductive biases, regularize the model and improve performance -- especially in the presence of limited data. In this work, we propose a simple and efficient way for optimizing hyperparameters inspired by the marginal likelihood, an optimization objective that requires no validation data. Our method partitions the training data and a neural network model into $K$ data shards and parameter partitions, respectively. Each partition is associated with and optimized only on specific data shards. Combining these partitions into subnetworks allows us to define the ``out-of-training-sample" loss of a subnetwork, i.e., the loss on data shards unseen by the subnetwork, as the objective for hyperparameter optimization. We demonstrate that we can apply this objective to optimize a variety of different hyperparameters in a single training run while being significantly computationally cheaper than alternative methods aiming to optimize the marginal likelihood for neural networks. Lastly, we also focus on optimizing hyperparameters in federated learning, where retraining and cross-validation are particularly challenging.

Abstract: Simulation-free methods for training continuous-time generative models construct probability paths that go between noise distributions and individual data samples. Recent works, such as Flow Matching, derived paths that are optimal for each data sample. However, these algorithms rely on independent data and noise samples, and do not exploit underlying structure in the data distribution for constructing probability paths. We propose Multisample Flow Matching, a more general framework that uses non-trivial couplings between data and noise samples while satisfying the correct marginal constraints. At very small overhead costs, this generalization allows us to (i) reduce gradient variance during training, (ii) obtain straighter flows for the learned vector field, which allows us to generate high-quality samples using fewer function evaluations, and (iii) obtain transport maps with lower cost in high dimensions, which has applications beyond generative modeling. Importantly, we do so in a completely simulation-free manner with a simple minimization objective. We show that our proposed methods improve sample consistency on downsampled ImageNet data sets, and lead to better low-cost sample generation.

Abstract: Graph Neural Networks (GNNs) are a form of deep learning that enable a wide range of machine learning applications on graph-structured data. The learning of GNNs, however, is known to pose challenges for memory-constrained devices such as GPUs. In this paper, we study exact compression as a way to reduce the memory requirements of learning GNNs on large graphs. In particular, we adopt a formal approach to compression and propose a methodology that transforms GNN learning problems into provably equivalent compressed GNN learning problems. In a preliminary experimental evaluation, we give insights into the compression ratios that can be obtained on real-world graphs and apply our methodology to an existing GNN benchmark.

Abstract: In high performance systems it is sometimes hard to build very large graphs that are efficient both with respect to memory and compute. This paper proposes a data structure called Markov-chain-priority-queue (MCPrioQ), which is a lock-free sparse markov-chain that enables online and continuous learning with time-complexity of $O(1)$ for updates and $O(CDF^{-1}(t))$ inference. MCPrioQ is especially suitable for recommender-systems for lookups of $n$-items in descending probability order. The concurrent updates are achieved using hash-tables and atomic instructions and the lookups are achieved through a novel priority-queue which allows for approximately correct results even during concurrent updates. The approximatly correct and lock-free property is maintained by a read-copy-update scheme, but where the semantics have been slightly updated to allow for swap of elements rather than the traditional pop-insert scheme.

Abstract: To stay competitive in the growing dairy market, farmers must continuously improve their livestock production systems. Precision livestock farming technologies provide individualised monitoring of animals on commercial farms, optimising livestock production. Continuous acoustic monitoring is a widely accepted sensing technique used to estimate the daily rumination and grazing time budget of free-ranging cattle. However, typical environmental and natural noises on pasture noticeably affect the performance and generalisation of current acoustic methods. In this study, we present an acoustic method called Noise-Robust Foraging Activity Recognizer (NRFAR). The proposed method determines foraging activity bouts by analysing fixed-length segments of identified jaw movement events associated with grazing and rumination. The additive noise robustness of NRFAR was evaluated for several signal-to-noise ratios, using stationary Gaussian white noise and four different non-stationary natural noise sources. In noiseless conditions, NRFAR reaches an average balanced accuracy of 89%, outperforming two previous acoustic methods by more than 7%. Additionally, NRFAR presents better performance than previous acoustic methods in 66 out of 80 evaluated noisy scenarios (p<0.01). NRFAR operates online with a similar computational cost to previous acoustic methods. The combination of these properties and the high performance in harsh free-ranging environments render NRFAR an excellent choice for real-time implementation in a low-power embedded device. The instrumentation and computational algorithms presented within this publication are protected by a pending patent application: AR P20220100910. Web demo available at: https://sinc.unl.edu.ar/web-demo/nrfar

Abstract: Many machine learning applications encounter a situation where model providers are required to further refine the previously trained model so as to gratify the specific need of local users. This problem is reduced to the standard model tuning paradigm if the target data is permissibly fed to the model. However, it is rather difficult in a wide range of practical cases where target data is not shared with model providers but commonly some evaluations about the model are accessible. In this paper, we formally set up a challenge named \emph{Earning eXtra PerformancE from restriCTive feEDdbacks} (EXPECTED) to describe this form of model tuning problems. Concretely, EXPECTED admits a model provider to access the operational performance of the candidate model multiple times via feedback from a local user (or a group of users). The goal of the model provider is to eventually deliver a satisfactory model to the local user(s) by utilizing the feedbacks. Unlike existing model tuning methods where the target data is always ready for calculating model gradients, the model providers in EXPECTED only see some feedbacks which could be as simple as scalars, such as inference accuracy or usage rate. To enable tuning in this restrictive circumstance, we propose to characterize the geometry of the model performance with regard to model parameters through exploring the parameters' distribution. In particular, for the deep models whose parameters distribute across multiple layers, a more query-efficient algorithm is further tailor-designed that conducts layerwise tuning with more attention to those layers which pay off better. Our theoretical analyses justify the proposed algorithms from the aspects of both efficacy and efficiency. Extensive experiments on different applications demonstrate that our work forges a sound solution to the EXPECTED problem.

Abstract: This paper presents a computational framework for the concise encoding of an ensemble of persistence diagrams, in the form of weighted Wasserstein barycenters [99], [101] of a dictionary of atom diagrams. We introduce a multi-scale gradient descent approach for the efficient resolution of the corresponding minimization problem, which interleaves the optimization of the barycenter weights with the optimization of the atom diagrams. Our approach leverages the analytic expressions for the gradient of both sub-problems to ensure fast iterations and it additionally exploits shared-memory parallelism. Extensive experiments on public ensembles demonstrate the efficiency of our approach, with Wasserstein dictionary computations in the orders of minutes for the largest examples. We show the utility of our contributions in two applications. First, we apply Wassserstein dictionaries to data reduction and reliably compress persistence diagrams by concisely representing them with their weights in the dictionary. Second, we present a dimensionality reduction framework based on a Wasserstein dictionary defined with a small number of atoms (typically three) and encode the dictionary as a low dimensional simplex embedded in a visual space (typically in 2D). In both applications, quantitative experiments assess the relevance of our framework. Finally, we provide a C++ implementation that can be used to reproduce our results.

Abstract: TADS are a novel, concise white-box representation of neural networks. In this paper, we apply TADS to the problem of neural network verification, using them to generate either proofs or concise error characterizations for desirable neural network properties. In a case study, we consider the robustness of neural networks to adversarial attacks, i.e., small changes to an input that drastically change a neural networks perception, and show that TADS can be used to provide precise diagnostics on how and where robustness errors a occur. We achieve these results by introducing Precondition Projection, a technique that yields a TADS describing network behavior precisely on a given subset of its input space, and combining it with PCA, a traditional, well-understood dimensionality reduction technique. We show that PCA is easily compatible with TADS. All analyses can be implemented in a straightforward fashion using the rich algebraic properties of TADS, demonstrating the utility of the TADS framework for neural network explainability and verification. While TADS do not yet scale as efficiently as state-of-the-art neural network verifiers, we show that, using PCA-based simplifications, they can still scale to mediumsized problems and yield concise explanations for potential errors that can be used for other purposes such as debugging a network or generating new training samples.

Abstract: The COVID 19 pandemic and ongoing political and regional conflicts have a highly detrimental impact on the global supply chain, causing significant delays in logistics operations and international shipments. One of the most pressing concerns is the uncertainty surrounding the availability dates of products, which is critical information for companies to generate effective logistics and shipment plans. Therefore, accurately predicting availability dates plays a pivotal role in executing successful logistics operations, ultimately minimizing total transportation and inventory costs. We investigate the prediction of product availability dates for General Electric (GE) Gas Power's inbound shipments for gas and steam turbine service and manufacturing operations, utilizing both numerical and categorical features. We evaluate several regression models, including Simple Regression, Lasso Regression, Ridge Regression, Elastic Net, Random Forest (RF), Gradient Boosting Machine (GBM), and Neural Network models. Based on real world data, our experiments demonstrate that the tree based algorithms (i.e., RF and GBM) provide the best generalization error and outperforms all other regression models tested. We anticipate that our prediction models will assist companies in managing supply chain disruptions and reducing supply chain risks on a broader scale.

Abstract: The field of machine learning (ML) has gained widespread adoption, leading to a significant demand for adapting ML to specific scenarios, which is yet expensive and non-trivial. The predominant approaches towards the automation of solving ML tasks (e.g., AutoML) are often time consuming and hard to understand for human developers. In contrast, though human engineers have the incredible ability to understand tasks and reason about solutions, their experience and knowledge are often sparse and difficult to utilize by quantitative approaches. In this paper, we aim to bridge the gap between machine intelligence and human knowledge by introducing a novel framework MLCopilot, which leverages the state-of-the-art LLMs to develop ML solutions for novel tasks. We showcase the possibility of extending the capability of LLMs to comprehend structured inputs and perform thorough reasoning for solving novel ML tasks. And we find that, after some dedicated design, the LLM can (i) observe from the existing experiences of ML tasks and (ii) reason effectively to deliver promising results for new tasks. The solution generated can be used directly to achieve high levels of competitiveness.

Abstract: Federated learning has shown enormous promise as a way of training ML models in distributed environments while reducing communication costs and protecting data privacy. However, the rise of complex cyber-physical systems, such as the Internet-of-Things, presents new challenges that are not met with traditional FL methods. Hierarchical Federated Learning extends the traditional FL process to enable more efficient model aggregation based on application needs or characteristics of the deployment environment (e.g., resource capabilities and/or network connectivity). It illustrates the benefits of balancing processing across the cloud-edge continuum. Hierarchical Federated Learning is likely to be a key enabler for a wide range of applications, such as smart farming and smart energy management, as it can improve performance and reduce costs, whilst also enabling FL workflows to be deployed in environments that are not well-suited to traditional FL. Model aggregation algorithms, software frameworks, and infrastructures will need to be designed and implemented to make such solutions accessible to researchers and engineers across a growing set of domains. H-FL also introduces a number of new challenges. For instance, there are implicit infrastructural challenges. There is also a trade-off between having generalised models and personalised models. If there exist geographical patterns for data (e.g., soil conditions in a smart farm likely are related to the geography of the region itself), then it is crucial that models used locally can consider their own locality in addition to a globally-learned model. H-FL will be crucial to future FL solutions as it can aggregate and distribute models at multiple levels to optimally serve the trade-off between locality dependence and global anomaly robustness.

Abstract: We study $K$-armed bandit problems where the reward distributions of the arms are all supported on the $[0,1]$ interval. It has been a challenge to design regret-efficient randomized exploration algorithms in this setting. Maillard sampling~\cite{maillard13apprentissage}, an attractive alternative to Thompson sampling, has recently been shown to achieve competitive regret guarantees in the sub-Gaussian reward setting~\cite{bian2022maillard} while maintaining closed-form action probabilities, which is useful for offline policy evaluation. In this work, we propose the Kullback-Leibler Maillard Sampling (KL-MS) algorithm, a natural extension of Maillard sampling for achieving KL-style gap-dependent regret bound. We show that KL-MS enjoys the asymptotic optimality when the rewards are Bernoulli and has a worst-case regret bound of the form $O(\sqrt{\mu^*(1-\mu^*) K T \ln K} + K \ln T)$, where $\mu^*$ is the expected reward of the optimal arm, and $T$ is the time horizon length.

Abstract: Unlike conventional grid and mesh based methods for solving partial differential equations (PDEs), neural networks have the potential to break the curse of dimensionality, providing approximate solutions to problems where using classical solvers is difficult or impossible. While global minimization of the PDE residual over the network parameters works well for boundary value problems, catastrophic forgetting impairs the applicability of this approach to initial value problems (IVPs). In an alternative local-in-time approach, the optimization problem can be converted into an ordinary differential equation (ODE) on the network parameters and the solution propagated forward in time; however, we demonstrate that current methods based on this approach suffer from two key issues. First, following the ODE produces an uncontrolled growth in the conditioning of the problem, ultimately leading to unacceptably large numerical errors. Second, as the ODE methods scale cubically with the number of model parameters, they are restricted to small neural networks, significantly limiting their ability to represent intricate PDE initial conditions and solutions. Building on these insights, we develop Neural IVP, an ODE based IVP solver which prevents the network from getting ill-conditioned and runs in time linear in the number of parameters, enabling us to evolve the dynamics of challenging PDEs with neural networks.

Abstract: Recent work in mechanistic interpretability has reverse-engineered nontrivial behaviors of transformer models. These contributions required considerable effort and researcher intuition, which makes it difficult to apply the same methods to understand the complex behavior that current models display. At their core however, the workflow for these discoveries is surprisingly similar. Researchers create a data set and metric that elicit the desired model behavior, subdivide the network into appropriate abstract units, replace activations of those units to identify which are involved in the behavior, and then interpret the functions that these units implement. By varying the data set, metric, and units under investigation, researchers can understand the functionality of each neural network region and the circuits they compose. This work proposes a novel algorithm, Automatic Circuit DisCovery (ACDC), to automate the identification of the important units in the network. Given a model's computational graph, ACDC finds subgraphs that explain a behavior of the model. ACDC was able to reproduce a previously identified circuit for Python docstrings in a small transformer, identifying 6/7 important attention heads that compose up to 3 layers deep, while including 91% fewer the connections.