Search

The availability of continuous glucose monitors as over-the-counter commodities have created a unique opportunity to monitor a person's blood glucose levels, forecast blood glucose trajectories and provide automated interventions to prevent devastating chronic complications that arise from poor glucose control. However, forecasting blood glucose levels is challenging because blood glucose changes consistently in response to food intake, medication intake, physical activity, sleep, and stress. It is particularly difficult to accurately predict BGL from multimodal and irregularly sampled data and over long prediction horizons. Furthermore, these forecasting models must operate in real-time on edge devices to provide in-the-moment interventions. To address these challenges, we propose GlucoNet, an AI-powered sensor system for continuously monitoring behavioral and physiological health and robust forecasting of blood glucose patterns. GlucoNet devises a feature decomposition-based transformer model that incorporates patients' behavioral and physiological data and transforms sparse and irregular patient data (e.g., diet and medication intake data) into continuous features using a mathematical model, facilitating better integration with the BGL data. Given the non-linear and non-stationary nature of BG signals, we propose a decomposition method to extract both low and high-frequency components from the BGL signals, thus providing accurate forecasting. To reduce the computational complexity, we also propose to employ knowledge distillation to compress the transformer model. GlucoNet achieves a 60% improvement in RMSE and a 21% reduction in the number of parameters, using data obtained involving 12 participants with T1-Diabetes. These results underscore GlucoNet's potential as a compact and reliable tool for real-world diabetes prevention and management.

Metaheuristics are stochastic optimization algorithms that mimic natural processes to find optimal solutions to complex problems. The success of metaheuristics largely depends on the ability to effectively explore and exploit the search space. Memory mechanisms have been introduced in several popular metaheuristic algorithms to enhance their performance. This chapter explores the significance of memory in metaheuristic algorithms and provides insights from well-known algorithms. The chapter begins by introducing the concept of memory, and its role in metaheuristic algorithms. The key factors influencing the effectiveness of memory mechanisms are discussed, such as the size of the memory, the information stored in memory, and the rate of information decay. A comprehensive analysis of how memory mechanisms are incorporated into popular metaheuristic algorithms is presented and concludes by highlighting the importance of memory in metaheuristic performance and providing future research directions for improving memory mechanisms. The key takeaways are that memory mechanisms can significantly enhance the performance of metaheuristics by enabling them to explore and exploit the search space effectively and efficiently, and that the choice of memory mechanism should be tailored to the problem domain and the characteristics of the search space., Comment: 25 pages, 5 figures, book chapter, Springer

An improved bilinear fuzzy genetic algorithm (BFGA) is introduced in this chapter for the design optimization of steel structures with semi-rigid connections. Semi-rigid connections provide a compromise between the stiffness of fully rigid connections and the flexibility of fully pinned connections. However, designing such structures is challenging due to the nonlinear behavior of semi-rigid connections. The BFGA is a robust optimization method that combines the strengths of fuzzy logic and genetic algorithm to handle the complexity and uncertainties of structural design problems. The BFGA, compared to standard GA, demonstrated to generate high-quality solutions in a reasonable time. The application of the BFGA is demonstrated through the optimization of steel structures with semirigid connections, considering the weight and performance criteria. The results show that the proposed BFGA is capable of finding optimal designs that satisfy all the design requirements and constraints. The proposed approach provides a promising solution for the optimization of complex structures with nonlinear behavior., Comment: 19 pages, 12 figures, book chapter, Springer

Opposition-based learning (OBL) is an effective approach to improve the performance of metaheuristic optimization algorithms, which are commonly used for solving complex engineering problems. This chapter provides a comprehensive review of the literature on the use of opposition strategies in metaheuristic optimization algorithms, discussing the benefits and limitations of this approach. An overview of the opposition strategy concept, its various implementations, and its impact on the performance of metaheuristic algorithms are presented. Furthermore, case studies on the application of opposition strategies in engineering problems are provided, including the optimum locating of control systems in tall building. A shear frame with Magnetorheological (MR) fluid damper is considered as a case study. The results demonstrate that the incorporation of opposition strategies in metaheuristic algorithms significantly enhances the quality and speed of the optimization process. This chapter aims to provide a clear understanding of the opposition strategy in metaheuristic optimization algorithms and its engineering applications, with the ultimate goal of facilitating its adoption in real-world engineering problems., Comment: 17 pages, 4 figures, book chapter, Springer

Building deep reinforcement learning (RL) agents that find a good policy with few samples has proven notoriously challenging. To achieve sample efficiency, recent work has explored updating neural networks with large numbers of gradient steps for every new sample. While such high update-to-data (UTD) ratios have shown strong empirical performance, they also introduce instability to the training process. Previous approaches need to rely on periodic neural network parameter resets to address this instability, but restarting the training process is infeasible in many real-world applications and requires tuning the resetting interval. In this paper, we focus on one of the core difficulties of stable training with limited samples: the inability of learned value functions to generalize to unobserved on-policy actions. We mitigate this issue directly by augmenting the off-policy RL training process with a small amount of data generated from a learned world model. Our method, Model-Augmented Data for Temporal Difference learning (MAD-TD) uses small amounts of generated data to stabilize high UTD training and achieve competitive performance on the most challenging tasks in the DeepMind control suite. Our experiments further highlight the importance of employing a good model to generate data, MAD-TD's ability to combat value overestimation, and its practical stability gains for continued learning.

Existing 3D facial emotion modeling have been constrained by limited emotion classes and insufficient datasets. This paper introduces "Emo3D", an extensive "Text-Image-Expression dataset" spanning a wide spectrum of human emotions, each paired with images and 3D blendshapes. Leveraging Large Language Models (LLMs), we generate a diverse array of textual descriptions, facilitating the capture of a broad spectrum of emotional expressions. Using this unique dataset, we conduct a comprehensive evaluation of language-based models' fine-tuning and vision-language models like Contranstive Language Image Pretraining (CLIP) for 3D facial expression synthesis. We also introduce a new evaluation metric for this task to more directly measure the conveyed emotion. Our new evaluation metric, Emo3D, demonstrates its superiority over Mean Squared Error (MSE) metrics in assessing visual-text alignment and semantic richness in 3D facial expressions associated with human emotions. "Emo3D" has great applications in animation design, virtual reality, and emotional human-computer interaction., Comment: 11 pages, 10 figures

Critical Casimir forces emerge among particles or surfaces immersed in a near-critical fluid, with the sign of the force determined by surface properties and with its strength tunable by minute temperature changes. Here, we show how such forces can be used to trap a colloidal particle and levitate it above a substrate with a bull's-eye pattern consisting of a ring with surface properties opposite to the rest of the substrate. Using the Derjaguin approximation and mean-field calculations, we find a rich behavior of spherical colloids at such a patterned surface, including sedimentation towards the ring and levitation above the ring (ring levitation) or above the bull's-eye's center (point levitation). Within the Derjaguin approximation, we calculate a levitation diagram for point levitation showing the depth of the trapping potential and the height at which the colloid levitates, both depending on the pattern properties, the colloid size, and the solution temperature. Our calculations reveal that the parameter space associated with point levitation shrinks if the system is driven away from a critical point, while, surprisingly, the trapping force becomes stronger. We discuss the application of critical Casimir levitation for sorting colloids by size and for determining the thermodynamic distance to criticality. Our results show that critical Casimir forces provide rich opportunities for controlling the behavior of colloidal particles at patterned surfaces.

The Value Iteration (VI) algorithm is an iterative procedure to compute the value function of a Markov decision process, and is the basis of many reinforcement learning (RL) algorithms as well. As the error convergence rate of VI as a function of iteration $k$ is $O(\gamma^k)$, it is slow when the discount factor $\gamma$ is close to $1$. To accelerate the computation of the value function, we propose Deflated Dynamics Value Iteration (DDVI). DDVI uses matrix splitting and matrix deflation techniques to effectively remove (deflate) the top $s$ dominant eigen-structure of the transition matrix $\mathcal{P}^{\pi}$. We prove that this leads to a $\tilde{O}(\gamma^k |\lambda_{s+1}|^k)$ convergence rate, where $\lambda_{s+1}$is $(s+1)$-th largest eigenvalue of the dynamics matrix. We then extend DDVI to the RL setting and present Deflated Dynamics Temporal Difference (DDTD) algorithm. We empirically show the effectiveness of the proposed algorithms.

Long-horizon tasks, which have a large discount factor, pose a challenge for most conventional reinforcement learning (RL) algorithms. Algorithms such as Value Iteration and Temporal Difference (TD) learning have a slow convergence rate and become inefficient in these tasks. When the transition distributions are given, PID VI was recently introduced to accelerate the convergence of Value Iteration using ideas from control theory. Inspired by this, we introduce PID TD Learning and PID Q-Learning algorithms for the RL setting, in which only samples from the environment are available. We give a theoretical analysis of the convergence of PID TD Learning and its acceleration compared to the conventional TD Learning. We also introduce a method for adapting PID gains in the presence of noise and empirically verify its effectiveness.

We investigate the impact of auxiliary learning tasks such as observation reconstruction and latent self-prediction on the representation learning problem in reinforcement learning. We also study how they interact with distractions and observation functions in the MDP. We provide a theoretical analysis of the learning dynamics of observation reconstruction, latent self-prediction, and TD learning in the presence of distractions and observation functions under linear model assumptions. With this formalization, we are able to explain why latent-self prediction is a helpful \emph{auxiliary task}, while observation reconstruction can provide more useful features when used in isolation. Our empirical analysis shows that the insights obtained from our learning dynamics framework predicts the behavior of these loss functions beyond the linear model assumption in non-linear neural networks. This reinforces the usefulness of the linear model framework not only for theoretical analysis, but also practical benefit for applied problems.

Deep Neural Networks (DNNs) have advanced in many real-world applications, such as healthcare and autonomous driving. However, their high computational complexity and vulnerability to adversarial attacks are ongoing challenges. In this letter, approximate multipliers are used to explore DNN robustness improvement against adversarial attacks. By uniformly replacing accurate multipliers for state-of-the-art approximate ones in DNN layer models, we explore the DNNs robustness against various adversarial attacks in a feasible time. Results show up to 7% accuracy drop due to approximations when no attack is present while improving robust accuracy up to 10% when attacks applied.

We show that deep reinforcement learning algorithms can retain their ability to learn without resetting network parameters in settings where the number of gradient updates greatly exceeds the number of environment samples by combatting value function divergence. Under large update-to-data ratios, a recent study by Nikishin et al. (2022) suggested the emergence of a primacy bias, in which agents overfit early interactions and downplay later experience, impairing their ability to learn. In this work, we investigate the phenomena leading to the primacy bias. We inspect the early stages of training that were conjectured to cause the failure to learn and find that one fundamental challenge is a long-standing acquaintance: value function divergence. Overinflated Q-values are found not only on out-of-distribution but also in-distribution data and can be linked to overestimation on unseen action prediction propelled by optimizer momentum. We employ a simple unit-ball normalization that enables learning under large update ratios, show its efficacy on the widely used dm_control suite, and obtain strong performance on the challenging dog tasks, competitive with model-based approaches. Our results question, in parts, the prior explanation for sub-optimal learning due to overfitting early data., Comment: Accepted as a conference paper at the First Reinforcement Learning Conference (RLC)

Instantons present a deep insight into non-perturbative effects both in physics and mathematics. While leading instanton effects can be calculated simply as an exponent of the instanton action, the calculation of subleading contributions usually requires the spectrum of fluctuation operator on the instanton background and its Green's function, explicit knowledge of which is rare and a great success. Thus, we propose an inverse problem, namely, the reconstruction of the nonlinear action of the theory admitting instantons from the given fluctuation operator with a known Green's function. We constructively build the solution for this problem and apply it to a wide class of exactly solvable Schr\"{o}dinger operators, called shape-invariant operators, and its simpler subclass, namely reflectionless P\"{o}schl-Teller operators. In the latter case, we found that for the most values of parameters the reconstructed potentials are naturally defined not on the real line, but on some special multisheet covering of the complex plane, and discuss its physical interpretation. For the wider but less simple class of shape-invariant operators, we derive the set of parameters leading to the new infinite families of analytic potentials., Comment: 30+9 pages, 13 figures; comments are welcome

In this paper, $F(\nu)$ cosmology is proposed for the accelerating universe with asymptotic de Sitter expansion in terms of Hankel function index $\nu$. To some extent, both the initial expansion during early inflation and the current accelerated expansion can be studied with a vacuum cosmic fluid i.e. $\Lambda$ in the pure de Sitter phase. Observational data further support the notion of a quasi-vacuum fluid, rather than a pure vacuum, contributing to the quasi-de Sitter acceleration in both the early and late universe. By examining the asymptotic expansion of the Henkel function as an approximate solution of the Mukhanov-Sasaki equation, we seek a more detailed study of quasi-de Sitter solutions in cosmology containing vacuum-like fluid., Comment: 11 pages, 2 tables

Detecting faults in steel plates is crucial for ensuring the safety and reliability of the structures and industrial equipment. Early detection of faults can prevent further damage and costly repairs. This chapter aims at diagnosing and predicting the likelihood of steel plates developing faults using experimental text data. Various machine learning methods such as GWO-based and FDO-based MLP and CMLP are tested to classify steel plates as either faulty or non-faulty. The experiments produced promising results for all models, with similar accuracy and performance. However, the FDO-based MLP and CMLP models consistently achieved the best results, with 100% accuracy in all tested datasets. The other models' outcomes varied from one experiment to another. The findings indicate that models that employed the FDO as a learning algorithm had the potential to achieve higher accuracy with a little longer runtime compared to other algorithms. In conclusion, early detection of faults in steel plates is critical for maintaining safety and reliability, and machine learning techniques can help predict and diagnose these faults accurately., Comment: 19 pages

The manipulation of microscopic objects requires precise and controllable forces and torques. Recent advances have led to the use of critical Casimir forces as a powerful tool, which can be finely tuned through the temperature of the environment and the chemical properties of the involved objects. For example, these forces have been used to self-organize ensembles of particles and to counteract stiction caused by Casimir-Liftshitz forces. However, until now, the potential of critical Casimir torques has been largely unexplored. Here, we demonstrate that critical Casimir torques can efficiently control the alignment of microscopic objects on nanopatterned substrates. We show experimentally and corroborate with theoretical calculations and Monte Carlo simulations that circular patterns on a substrate can stabilize the position and orientation of microscopic disks. By making the patterns elliptical, such microdisks can be subject to a torque which flips them upright while simultaneously allowing for more accurate control of the microdisk position. More complex patterns can selectively trap 2D-chiral particles and generate particle motion similar to non-equilibrium Brownian ratchets. These findings provide new opportunities for nanotechnological applications requiring precise positioning and orientation of microscopic objects.

Neural networks are susceptible to adversarial perturbations that are transferable across different models. In this paper, we introduce a novel model alignment technique aimed at improving a given source model's ability in generating transferable adversarial perturbations. During the alignment process, the parameters of the source model are fine-tuned to minimize an alignment loss. This loss measures the divergence in the predictions between the source model and another, independently trained model, referred to as the witness model. To understand the effect of model alignment, we conduct a geometric analysis of the resulting changes in the loss landscape. Extensive experiments on the ImageNet dataset, using a variety of model architectures, demonstrate that perturbations generated from aligned source models exhibit significantly higher transferability than those from the original source model., Comment: Accepted at the European Conference on Computer Vision (ECCV) 2024. Code: https://github.com/averyma/model-alignment

Stochastic gradient descent (SGD) and adaptive gradient methods, such as Adam and RMSProp, have been widely used in training deep neural networks. We empirically show that while the difference between the standard generalization performance of models trained using these methods is small, those trained using SGD exhibit far greater robustness under input perturbations. Notably, our investigation demonstrates the presence of irrelevant frequencies in natural datasets, where alterations do not affect models' generalization performance. However, models trained with adaptive methods show sensitivity to these changes, suggesting that their use of irrelevant frequencies can lead to solutions sensitive to perturbations. To better understand this difference, we study the learning dynamics of gradient descent (GD) and sign gradient descent (signGD) on a synthetic dataset that mirrors natural signals. With a three-dimensional input space, the models optimized with GD and signGD have standard risks close to zero but vary in their adversarial risks. Our result shows that linear models' robustness to $\ell_2$-norm bounded changes is inversely proportional to the model parameters' weight norm: a smaller weight norm implies better robustness. In the context of deep learning, our experiments show that SGD-trained neural networks show smaller Lipschitz constants, explaining the better robustness to input perturbations than those trained with adaptive gradient methods.

We design a novel algorithm for optimal transport by drawing from the entropic optimal transport, mirror descent and conjugate gradients literatures. Our scalable and GPU parallelizable algorithm is able to compute the Wasserstein distance with extreme precision, reaching relative error rates of $10^{-8}$ without numerical stability issues. Empirically, the algorithm converges to high precision solutions more quickly in terms of wall-clock time than a variety of algorithms including log-domain stabilized Sinkhorn's Algorithm. We provide careful ablations with respect to algorithm and problem parameters, and present benchmarking over upsampled MNIST images, comparing to various recent algorithms over high-dimensional problems. The results suggest that our algorithm can be a useful addition to the practitioner's optimal transport toolkit.

We consider the problem of learning models for risk-sensitive reinforcement learning. We theoretically demonstrate that proper value equivalence, a method of learning models which can be used to plan optimally in the risk-neutral setting, is not sufficient to plan optimally in the risk-sensitive setting. We leverage distributional reinforcement learning to introduce two new notions of model equivalence, one which is general and can be used to plan for any risk measure, but is intractable; and a practical variation which allows one to choose which risk measures they may plan optimally for. We demonstrate how our framework can be used to augment any model-free risk-sensitive algorithm, and provide both tabular and large-scale experiments to demonstrate its ability.