Automated hyperparameter search in machine learning, especially for deep learning models, is typically formulated as a bilevel optimization problem, with hyperparameter values determined by the upper level and the model learning achieved by the lower-level problem. Most of the existing bilevel optimization solutions either assume the uniqueness of the optimal training model given hyperparameters or adopt an optimistic view when the non-uniqueness issue emerges. Potential model uncertainty may arise when training complex models with limited data, especially when the uniqueness assumption is violated. Thus, the suitability of the optimistic view underlying current bilevel hyperparameter optimization solutions is questionable. In this paper, we propose pessimistic bilevel hyperparameter optimization to assure appropriate outer-level hyperparameters to better generalize the inner-level learned models, by explicitly incorporating potential uncertainty of the inner-level solution set. To solve the resulting computationally challenging pessimistic bilevel optimization problem, we develop a novel relaxation-based approximation method. It derives pessimistic solutions with more robust prediction models. In our empirical studies of automated hyperparameter search for binary linear classifiers, pessimistic solutions have demonstrated better prediction performances than optimistic counterparts when we have limited training data or perturbed testing data, showing the necessity of considering pessimistic solutions besides existing optimistic ones.
On High-Performance Computing (HPC) systems, several hyperparameter configurations can be evaluated in parallel to speed up the Hyperparameter Optimization (HPO) process. State-of-the-art HPO methods follow a bandit-based approach and build on top of successive halving, where the final performance of a combination is estimated based on a lower than fully trained fidelity performance metric and more promising combinations are assigned more resources over time. Frequently, the number of epochs is treated as a resource, letting more promising combinations train longer. Another option is to use the number of workers as a resource and directly allocate more workers to more promising configurations via data-parallel training. This article proposes a novel Resource-Adaptive Successive Doubling Algorithm (RASDA), which combines a resource-adaptive successive doubling scheme with the plain Asynchronous Successive Halving Algorithm (ASHA). Scalability of this approach is shown on up to 1,024 Graphics Processing Units (GPUs) on modern HPC systems. It is applied to different types of Neural Networks (NNs) and trained on large datasets from the Computer Vision (CV), Computational Fluid Dynamics (CFD), and Additive Manufacturing (AM) domains, where performing more than one full training run is usually infeasible. Empirical results show that RASDA outperforms ASHA by a factor of up to 1.9 with respect to the runtime. At the same time, the solution quality of final ASHA models is maintained or even surpassed by the implicit batch size scheduling of RASDA. With RASDA, systematic HPO is applied to a terabyte-scale scientific dataset for the first time in the literature, enabling efficient optimization of complex models on massive scientific data. The implementation of RASDA is available on https://github.com/olympiquemarcel/rasda
Majority subspace clustering (SC) algorithms depend on one or more hyperparameters that need to be carefully tuned for the SC algorithms to achieve high clustering performance. Hyperparameter optimization (HPO) is often performed using grid-search, assuming that some labeled data is available. In some domains, such as medicine, this assumption does not hold true in many cases. One avenue of research focuses on developing SC algorithms that are inherently free of hyperparameters. For hyperparameters-dependent SC algorithms, one approach to label-independent HPO tuning is based on internal clustering quality metrics (if available), whose performance should ideally match that of external (label-dependent) clustering quality metrics. In this paper, we propose a novel approach to label-independent HPO that uses clustering quality metrics, such as accuracy (ACC) or normalized mutual information (NMI), that are computed based on pseudo-labels obtained from the SC algorithm across a predefined grid of hyperparameters. Assuming that ACC (or NMI) is a smooth function of hyperparameter values it is possible to select subintervals of hyperparameters. These subintervals are then iteratively further split into halves or thirds until a relative error criterion is satisfied. In principle, the hyperparameters of any SC algorithm can be tuned using the proposed method. We demonstrate this approach on several single- and multi-view SC algorithms, comparing the achieved performance with their oracle versions across six datasets representing digits, faces and objects. The proposed method typically achieves clustering performance that is 5% to 7% lower than that of the oracle versions. We also make our proposed method interpretable by visualizing subspace bases, which are estimated from the computed clustering partitions. This aids in the initial selection of the hyperparameter search space.
Learning dynamical models from data is not only fundamental but also holds great promise for advancing principle discovery, time-series prediction, and controller design. Among various approaches, Gaussian Process State-Space Models (GPSSMs) have recently gained significant attention due to their combination of flexibility and interpretability. However, for online learning, the field lacks an efficient method suitable for scenarios where prior information regarding data distribution and model function is limited. To address this issue, this paper proposes a recursive GPSSM method with adaptive capabilities for both operating domains and Gaussian process (GP) hyperparameters. Specifically, we first utilize first-order linearization to derive a Bayesian update equation for the joint distribution between the system state and the GP model, enabling closed-form and domain-independent learning. Second, an online selection algorithm for inducing points is developed based on informative criteria to achieve lightweight learning. Third, to support online hyperparameter optimization, we recover historical measurement information from the current filtering distribution. Comprehensive evaluations on both synthetic and real-world datasets demonstrate the superior accuracy, computational efficiency, and adaptability of our method compared to state-of-the-art online GPSSM techniques.
Drawing motivation from the manifold hypothesis, which posits that most high-dimensional data lies on or near low-dimensional manifolds, we apply manifold learning to the space of neural networks. We learn manifolds where datapoints are neural networks by introducing a distance between the hidden layer representations of the neural networks. These distances are then fed to the non-linear dimensionality reduction algorithm PHATE to create a manifold of neural networks. We characterize this manifold using features of the representation, including class separation, hierarchical cluster structure, spectral entropy, and topological structure. Our analysis reveals that high-performing networks cluster together in the manifold, displaying consistent embedding patterns across all these features. Finally, we demonstrate the utility of this approach for guiding hyperparameter optimization and neural architecture search by sampling from the manifold.
Building effective machine learning (ML) workflows to address complex tasks is a primary focus of the Automatic ML (AutoML) community and a critical step toward achieving artificial general intelligence (AGI). Recently, the integration of Large Language Models (LLMs) into ML workflows has shown great potential for automating and enhancing various stages of the ML pipeline. This survey provides a comprehensive and up-to-date review of recent advancements in using LLMs to construct and optimize ML workflows, focusing on key components encompassing data and feature engineering, model selection and hyperparameter optimization, and workflow evaluation. We discuss both the advantages and limitations of LLM-driven approaches, emphasizing their capacity to streamline and enhance ML workflow modeling process through language understanding, reasoning, interaction, and generation. Finally, we highlight open challenges and propose future research directions to advance the effective application of LLMs in ML workflows.
With fairness concerns gaining significant attention in Machine Learning (ML), several bias mitigation techniques have been proposed, often compared against each other to find the best method. These benchmarking efforts tend to use a common setup for evaluation under the assumption that providing a uniform environment ensures a fair comparison. However, bias mitigation techniques are sensitive to hyperparameter choices, random seeds, feature selection, etc., meaning that comparison on just one setting can unfairly favour certain algorithms. In this work, we show significant variance in fairness achieved by several algorithms and the influence of the learning pipeline on fairness scores. We highlight that most bias mitigation techniques can achieve comparable performance, given the freedom to perform hyperparameter optimization, suggesting that the choice of the evaluation parameters-rather than the mitigation technique itself-can sometimes create the perceived superiority of one method over another. We hope our work encourages future research on how various choices in the lifecycle of developing an algorithm impact fairness, and trends that guide the selection of appropriate algorithms.
Predicting spatio-temporal traffic flow presents significant challenges due to complex interactions between spatial and temporal factors. Existing approaches often address these dimensions in isolation, neglecting their critical interdependencies. In this paper, we introduce the Spatio-Temporal Unitized Model (STUM), a unified framework designed to capture both spatial and temporal dependencies while addressing spatio-temporal heterogeneity through techniques such as distribution alignment and feature fusion. It also ensures both predictive accuracy and computational efficiency. Central to STUM is the Adaptive Spatio-temporal Unitized Cell (ASTUC), which utilizes low-rank matrices to seamlessly store, update, and interact with space, time, as well as their correlations. Our framework is also modular, allowing it to integrate with various spatio-temporal graph neural networks through components such as backbone models, feature extractors, residual fusion blocks, and predictive modules to collectively enhance forecasting outcomes. Experimental results across multiple real-world datasets demonstrate that STUM consistently improves prediction performance with minimal computational cost. These findings are further supported by hyperparameter optimization, pre-training analysis, and result visualization. We provide our source code for reproducibility at https://anonymous.4open.science/r/STUM-E4F0.
In this study, we apply two pillars of Scientific Machine Learning: Neural Ordinary Differential Equations (Neural ODEs) and Universal Differential Equations (UDEs) to the Lotka Volterra Predator Prey Model, a fundamental ecological model describing the dynamic interactions between predator and prey populations. The Lotka-Volterra model is critical for understanding ecological dynamics, population control, and species interactions, as it is represented by a system of differential equations. In this work, we aim to uncover the underlying differential equations without prior knowledge of the system, relying solely on training data and neural networks. Using robust modeling in the Julia programming language, we demonstrate that both Neural ODEs and UDEs can be effectively utilized for prediction and forecasting of the Lotka-Volterra system. More importantly, we introduce the forecasting breakdown point: the time at which forecasting fails for both Neural ODEs and UDEs. We observe how UDEs outperform Neural ODEs by effectively recovering the underlying dynamics and achieving accurate forecasting with significantly less training data. Additionally, we introduce Gaussian noise of varying magnitudes (from mild to high) to simulate real-world data perturbations and show that UDEs exhibit superior robustness, effectively recovering the underlying dynamics even in the presence of noisy data, while Neural ODEs struggle with high levels of noise. Through extensive hyperparameter optimization, we offer insights into neural network architectures, activation functions, and optimizers that yield the best results. This study opens the door to applying Scientific Machine Learning frameworks for forecasting tasks across a wide range of ecological and scientific domains.
Protein engineering is important for biomedical applications, but conventional approaches are often inefficient and resource-intensive. While deep learning (DL) models have shown promise, their training or implementation into protein engineering remains challenging for biologists without specialized computational expertise. To address this gap, we propose AutoProteinEngine (AutoPE), an agent framework that leverages large language models (LLMs) for multimodal automated machine learning (AutoML) for protein engineering. AutoPE innovatively allows biologists without DL backgrounds to interact with DL models using natural language, lowering the entry barrier for protein engineering tasks. Our AutoPE uniquely integrates LLMs with AutoML to handle model selection for both protein sequence and graph modalities, automatic hyperparameter optimization, and automated data retrieval from protein databases. We evaluated AutoPE through two real-world protein engineering tasks, demonstrating substantial performance improvements compared to traditional zero-shot and manual fine-tuning approaches. By bridging the gap between DL and biologists' domain expertise, AutoPE empowers researchers to leverage DL without extensive programming knowledge. Our code is available at https://github.com/tsynbio/AutoPE.
Conventional methods for Bayesian optimization (BO) primarily involve one-step optimal decisions (e.g., maximizing expected improvement of the next step). To avoid myopic behavior, multi-step lookahead BO algorithms such as rollout strategies consider the sequential decision-making nature of BO, i.e., as a stochastic dynamic programming (SDP) problem, demonstrating promising results in recent years. However, owing to the curse of dimensionality, most of these methods make significant approximations or suffer scalability issues, e.g., being limited to two-step lookahead. This paper presents a novel reinforcement learning (RL)-based framework for multi-step lookahead BO in high-dimensional black-box optimization problems. The proposed method enhances the scalability and decision-making quality of multi-step lookahead BO by efficiently solving the SDP of the BO process in a near-optimal manner using RL. We first introduce an Attention-DeepSets encoder to represent the state of knowledge to the RL agent and employ off-policy learning to accelerate its initial training. We then propose a multi-task, fine-tuning procedure based on end-to-end (encoder-RL) on-policy learning. We evaluate the proposed method, EARL-BO (Encoder Augmented RL for Bayesian Optimization), on both synthetic benchmark functions and real-world hyperparameter optimization problems, demonstrating significantly improved performance compared to existing multi-step lookahead and high-dimensional BO methods.
Multi-objective Bayesian optimization has been widely adopted in scientific experiment design, including drug discovery and hyperparameter optimization. In practice, regulatory or safety concerns often impose additional thresholds on certain attributes of the experimental outcomes. Previous work has primarily focused on constrained single-objective optimization tasks or active search under constraints. The existing constrained multi-objective algorithms address the issue with heuristics and approximations, posing challenges to the analysis of the sample efficiency. We propose a novel constrained multi-objective Bayesian optimization algorithm COMBOO that balances active learning of the level-set defined on multiple unknowns with multi-objective optimization within the feasible region. We provide both theoretical analysis and empirical evidence, demonstrating the efficacy of our approach on various synthetic benchmarks and real-world applications.
Model selection in Gaussian processes scales prohibitively with the size of the training dataset, both in time and memory. While many approximations exist, all incur inevitable approximation error. Recent work accounts for this error in the form of computational uncertainty, which enables -- at the cost of quadratic complexity -- an explicit tradeoff between computation and precision. Here we extend this development to model selection, which requires significant enhancements to the existing approach, including linear-time scaling in the size of the dataset. We propose a novel training loss for hyperparameter optimization and demonstrate empirically that the resulting method can outperform SGPR, CGGP and SVGP, state-of-the-art methods for GP model selection, on medium to large-scale datasets. Our experiments show that model selection for computation-aware GPs trained on 1.8 million data points can be done within a few hours on a single GPU. As a result of this work, Gaussian processes can be trained on large-scale datasets without significantly compromising their ability to quantify uncertainty -- a fundamental prerequisite for optimal decision-making.
Hyperparameters are configuration variables controlling the behavior of machine learning algorithms. They are ubiquitous in machine learning and artificial intelligence and the choice of their values determines the effectiveness of systems based on these technologies. Manual hyperparameter search is often unsatisfactory and becomes infeasible when the number of hyperparameters is large. Automating the search is an important step towards advancing, streamlining, and systematizing machine learning, freeing researchers and practitioners alike from the burden of finding a good set of hyperparameters by trial and error. In this survey, we present a unified treatment of hyperparameter optimization, providing the reader with examples, insights into the state-of-the-art, and numerous links to further reading. We cover the main families of techniques to automate hyperparameter search, often referred to as hyperparameter optimization or tuning, including random and quasi-random search, bandit-, model-, population-, and gradient-based approaches. We further discuss extensions, including online, constrained, and multi-objective formulations, touch upon connections with other fields such as meta-learning and neural architecture search, and conclude with open questions and future research directions.
This study introduces SLLMBO, an innovative framework leveraging large language models (LLMs) for hyperparameter optimization (HPO), incorporating dynamic search space adaptability, enhanced parameter space exploitation, and a novel LLM-tree-structured parzen estimator (LLM-TPE) sampler. By addressing limitations in recent fully LLM-based methods and traditional bayesian optimization (BO), SLLMBO achieves more robust optimization. This comprehensive benchmarking evaluates multiple LLMs, including GPT-3.5-Turbo, GPT-4o, Claude-Sonnet-3.5, and Gemini-1.5-Flash, extending prior work and establishing SLLMBO as the first framework to benchmark a diverse set of LLMs for HPO. By integrating LLMs' established strengths in parameter initialization with the exploitation abilities demonstrated in this study, alongside TPE's exploration capabilities, the LLM-TPE sampler achieves a balanced exploration-exploitation trade-off, reduces API costs, and mitigates premature early stoppings for more effective parameter searches. Across 14 tabular tasks in classification and regression, the LLM-TPE sampler outperformed fully LLM-based methods and achieved superior results over BO methods in 9 tasks. Testing early stopping in budget-constrained scenarios demonstrated competitive performance, indicating that LLM-based methods generally benefit from extended iterations for optimal results. This work lays the foundation for future research exploring open-source LLMs, reproducibility of LLM results in HPO, and benchmarking SLLMBO on complex datasets, such as image classification, segmentation, and machine translation.
The physicochemical properties of chemical compounds have great importance in several areas, including pharmaceuticals, environmental and separation science. Among these are physicochemical properties such as the octanol-water partition coefficient, which has been considered an important index pointing out lipophilicity and hydrophilicity. It affects drug absorption and membrane permeability. Following Lipinski's rule of five, logP was identified as one of the key determinants of the stability of chemical entities and, as such, needed state-of-the-art methods for measuring lipophilicity. This paper presents a deep-learning model, OWPCP, developed to compute logP using Morgan fingerprints and MACCS keys as input features. It uses the interconnection of such molecular representations with logP values extracted from 26,254 compounds. The dataset was prepared to contain a wide range of chemical structures with differing molecular weights and polar surface area. Hyperparameter optimization was conducted using the Keras Tuner alongside the Hyperband algorithm to enhance the performance. OWPCP demonstrated outstanding performance compared to current computational methods, achieving an MAE=0.247 on the test set and outperforming all previous DL models. Remarkably, while one of the most accurate recent models is based on experimental data on retention time to make predictions, OWPCP manages computing logP efficiently without depending on these factors, being, therefore, very useful during early-stage drug discovery. Our model outperforms the best model, which leverages Retention Time, and our model does not require any experimental data. Further validation of the model performance was done across different functional groups, and it showed very high accuracy, especially for compounds that contain aliphatic OH groups. The results have indicated that OWPCP provides a reliable prediction of logP.
Large-scale industrial recommendation models predict the most relevant items from catalogs containing millions or billions of options. To train these models efficiently, a small set of irrelevant items (negative samples) is selected from the vast catalog for each relevant item (positive example), helping the model distinguish between relevant and irrelevant items. Choosing the right negative sampling method is a common challenge. We address this by implementing and comparing various negative sampling methods - random, popularity-based, in-batch, mixed, adaptive, and adaptive with mixed variants - on modern sequential recommendation models. Our experiments, including hyperparameter optimization and 20x repeats on three benchmark datasets with varying popularity biases, show how the choice of method and dataset characteristics impact key model performance metrics. We also reveal that average performance metrics often hide imbalances across popularity bands (head, mid, tail). We find that commonly used random negative sampling reinforces popularity bias and performs best for head items. Popularity-based methods (in-batch and global popularity negative sampling) can offer balanced performance at the cost of lower overall model performance results. Our study serves as a practical guide to the trade-offs in selecting a negative sampling method for large-scale sequential recommendation models. Code, datasets, experimental results and hyperparameters are available at: https://github.com/apple/ml-negative-sampling.
We investigate the efficacy of data augmentations to close the domain gap in spaceborne computer vision, crucial for autonomous operations like on-orbit servicing. As the use of computer vision in space increases, challenges such as hostile illumination and low signal-to-noise ratios significantly hinder performance. While learning-based algorithms show promising results, their adoption is limited by the need for extensive annotated training data and the domain gap that arises from differences between synthesized and real-world imagery. This study explores domain generalization in terms of data augmentations -- classical color and geometric transformations, corruptions, and noise -- to enhance model performance across the domain gap. To this end, we conduct an large scale experiment using a hyperparameter optimization pipeline that samples hundreds of different configurations and searches for the best set to bridge the domain gap. As a reference task, we use 2D object detection and evaluate on the SPEED+ dataset that contains real hardware-in-the-loop satellite images in its test set. Moreover, we evaluate four popular object detectors, including Mask R-CNN, Faster R-CNN, YOLO-v7, and the open set detector GroundingDINO, and highlight their trade-offs between performance, inference speed, and training time. Our results underscore the vital role of data augmentations in bridging the domain gap, improving model performance, robustness, and reliability for critical space applications. As a result, we propose two novel data augmentations specifically developed to emulate the visual effects observed in orbital imagery. We conclude by recommending the most effective augmentations for advancing computer vision in challenging orbital environments. Code for training detectors and hyperparameter search will be made publicly available.
Accurate forecasting of electrical demand is essential for maintaining a stable and reliable power grid, optimizing the allocation of energy resources, and promoting efficient energy consumption practices. This study investigates the effectiveness of five hyperparameter optimization (HPO) algorithms -- Random Search, Covariance Matrix Adaptation Evolution Strategy (CMA--ES), Bayesian Optimization, Partial Swarm Optimization (PSO), and Nevergrad Optimizer (NGOpt) across univariate and multivariate Short-Term Load Forecasting (STLF) tasks. Using the Panama Electricity dataset (n=48,049), we evaluate HPO algorithms' performances on a surrogate forecasting algorithm, XGBoost, in terms of accuracy (i.e., MAPE, $R^2$) and runtime. Performance plots visualize these metrics across varying sample sizes from 1,000 to 20,000, and Kruskal--Wallis tests assess the statistical significance of the performance differences. Results reveal significant runtime advantages for HPO algorithms over Random Search. In univariate models, Bayesian optimization exhibited the lowest accuracy among the tested methods. This study provides valuable insights for optimizing XGBoost in the STLF context and identifies areas for future research.
In this study, we apply two pillars of Scientific Machine Learning: Neural Ordinary Differential Equations (Neural ODEs) and Universal Differential Equations (UDEs) to the Chandrasekhar White Dwarf Equation (CWDE). The CWDE is fundamental for understanding the life cycle of a star, and describes the relationship between the density of the white dwarf and its distance from the center. Despite the rise in Scientific Machine Learning frameworks, very less attention has been paid to the systematic applications of the above SciML pillars on astronomy based ODEs. Through robust modeling in the Julia programming language, we show that both Neural ODEs and UDEs can be used effectively for both prediction as well as forecasting of the CWDE. More importantly, we introduce the forecasting breakdown point - the time at which forecasting fails for both Neural ODEs and UDEs. Through a robust hyperparameter optimization testing, we provide insights on the neural network architecture, activation functions and optimizers which provide the best results. This study provides opens a door to investigate the applicability of Scientific Machine Learning frameworks in forecasting tasks for a wide range of scientific domains.