Tuning Time Research Articles

Predictive modeling and benchmarking for diamond price estimation: integrating classification, regression, hyperparameter tuning and execution time analysis

Predictive modeling and benchmarking for diamond price estimation: integrating classification, regression, hyperparameter tuning and execution time analysis

A novel bistable optical phase shifter using non-volatile MEMS actuator for application in photonic integrated circuits

Recently, micro-electromechanical systems (MEMS) actuators have been employed to perform the tuning mechanisms in photonic integrated circuits (PICs). Photonic components with MEMS-based tuning mechanisms offer several advantages, including small dimensions, low power consumption, fast tuning time, and high compatibility with silicon photonic platforms. These features enable the integration of photonic components on a single chip, which is demanded for complex applications such as artificial intelligence applications including machine learning, neural networks, and neuromorphic computing. Phase shifters are one of the most important components in PICs. This paper introduces a novel bistable optical phase shifter that utilizes a MEMS-based tuning mechanism to maintain its last state without consuming electrical power. The proposed device boasts impressive functional characteristics including a phase shift of 2.32π, a tuning time of 1.3μs, an optical loss of 0.03 dB, and compact dimensions of 35 μm by 35 μm while operating at a voltage of 18V. Given these remarkable features, the proposed device is an appealing candidate for integration in complex PICs such as photonics-based AI systems.

Dynamically Tunable Circularly Polarized Selectivity in Plasmon-Enhanced Halide Perovskite Nanocrystal Glasses.

Ultrafast spin manipulation for optical spin-logic applications requires material systems with strong spin-selective light-matter interactions. The optical Stark effect can realize spin-selective light-matter interactions by breaking the degeneracy of spin-selective transitions with an external electric field. Halide perovskites have large exciton binding energies, which enable a room-temperature optical Stark effect. However, halide perovskites are prone to degradation when interacting with light and polar solvents, limiting further integration with nanophotonic structures. We demonstrate a hybrid material system consisting of CsPbBr3 nanocrystal glass weakly coupled to resonant plasmonic silver nanoparticles, showing ultrafast tunable spin-based polarization selectivity at room temperature. We performed circularly polarized pump-probe characterizations to investigate the optical Stark effect in this material system, which resulted in a maximum energy shift of ∼3.67 meV (detuning energy of 0.11 eV and pump intensity of 0.62 GW/cm2). We show that halide perovskite nanocrystal glasses have excellent resistance to heat and moisture, which may be favorable for integration with nanophotonic structures for further engineering polarization states, energy tuning, and coherence time.

An interpretable ensemble machine-learning workflow for permeability predictions in tight sandstone reservoirs using logging data

Tight sandstone reservoirs exhibit strong vertical heterogeneity and complex pore structures, challenging conventional permeability evaluation methods based on well-logging data. Although rising machine-learning (ML) techniques have demonstrated excellent accuracy for industrial applications, the physics and rationality within such a powerful “black box” remain less clear. Hence, reliable permeability prediction would benefit from an interpretable ML-based workflow that could reveal the controlling factors. To compare the models and examine the underlying features, 16 different ML submodels are tested after data preprocessing, feature selection, and hyperparameter optimization. By comparing the fitting accuracy and tuning time, the light gradient boosting machine optimized by the whale optimization algorithm, referred to as LGB-WOA, is determined to be the optimal model with the best fitting accuracy and relatively short tuning time. A field data application demonstrates that even in highly heterogeneous reservoir sections, the LGB-WOA model outperformed conventional petrophysical models by being the most consistent with reservoir permeability directly measured from the core samples ([Formula: see text]). The Shapley additive explanation values are then used to interpret the predictions of our LGB-WOA model. As expected, the porosity curve exhibits the highest feature importance among all input features, significantly contributing to permeability predictions. Conversely, a wellbore diameter and compensated neutron log contribute the least and need not be used for subsequent model improvements. These experiments and workflow provide a powerful method for accurately assessing the permeability in complex reservoirs and contribute to a broader understanding of the application of ML in reservoir characterization, paving the way for establishing more interpretable and reliable prediction models.

Machine learning methods in evaluating the impact of economic factors on the consumer price index in Albania

The Consumer Price Index (CPI) in Albania is a measure of inflation that tracks changes in the prices of a basket of goods and services typically purchased by urban households in the country. It is a vital economic indicator used to assess changes in the cost of living and the overall price level in Albania. There are several factors that affect the levels and progress of the CPI, among them we have chosen: Euro/Lek and USD/Lek exchange rates, import levels, the monetary base, and salary data, from January 2007 to September 2023. In this paper, we investigate the efficiency of machine learning methods in determining the factors that have the greatest impact on the CPI. In our analysis, we assess the effectiveness of decision-tree models, Random Forest and XGBoost algorithms, in predicting the CPI behavior in Albania. Based on our empirical findings, we conclude that the monetary base and wages play a crucial role in influencing the CPI, with imports and exchange rates following closely in significance. Additionally, our results indicate that the Random Forest model demonstrates superior accuracy and demands less parameter tuning time compared to the alternatives. This research underscores the critical role of model selection in achieving precision and dependability in CPI forecasting. It underscores the immense potential of machine learning models in enhancing forecasting accuracy. The implications of this study are significant, as they can foster the creation of more precise and dependable forecasting models, equipping policymakers with a deeper understanding of economic stability.

Cross-architecture tuning of silicon and SiGe-based quantum devices using machine learning

The potential of Si and SiGe-based devices for the scaling of quantum circuits is tainted by device variability. Each device needs to be tuned to operation conditions and each device realisation requires a different tuning protocol. We demonstrate that it is possible to automate the tuning of a 4-gate Si FinFET, a 5-gate GeSi nanowire and a 7-gate Ge/SiGe heterostructure double quantum dot device from scratch with the same algorithm. We achieve tuning times of 30, 10, and 92 min, respectively. The algorithm also provides insight into the parameter space landscape for each of these devices, allowing for the characterization of the regions where double quantum dot regimes are found. These results show that overarching solutions for the tuning of quantum devices are enabled by machine learning.

Estimating the penetration rate of tunnel boring machines via gradient boosting algorithms

The prediction of tunnel boring machine (TBM) performance from the rate of penetration (ROP) point of view has yet to draw a lot of attention since it is one of the main challenges for excavation with TBMs. This study examined six tunnels excavated with TBM to develop predictive models of the ROP estimation using five algorithms: Gradient Boosting (GB), Extreme Gradient Boosting (XGBoost), Light Gradient Boosting Machine, Adaptive Boosting (AdaBoost), and CatBoost (categorical-features-include-GB). A dataset has been developed, including uniaxial compressive strength (UCS), Rock Type, the distance between planes of weakness (DPW), and thrust force (TF), and contains more than 575 data points for each parameter. The developed models showed that the XGBoost model outperformed the other models, followed by the CatBoost, according to seven different evaluation metrics used to rank the models when the models were modeled with the default values of the algorithm parameters. After tuning the hyperparameters, the GB model outperformed the others, while the other models remained relatively unchanged. By using the overall ranking according to the metrics and considering the parameter tuning time, XGBoost and CatBoost were presented as the two best models. SHAP (Shapley additive explanations: an explainable artificial intelligence tool) values and dependency plots showed that the TF has the highest impact on the ROP, followed by UCS, Rock Type, and DPW. It is concluded that the XGBoost and CatBoost models, with coefficients of determination of 0.9878 and 0.9682, respectively, may be used for estimating the TBM penetration rate for similar rock types.

Per-Instance Algorithm Configuration in Homogeneous Instance Spaces: A Use Case in Reconfigurable Assembly Systems

The physical capabilities of a reconfigurable assembly system (RAS) increase the agility and responsiveness of the system in highly volatile market conditions. However, achieving optimal RAS utilization entails solving complex optimization problems effectively and efficiently. These optimizations often define homogenous sets of problem instances. While algorithm configuration in such homogeneous contexts traditionally adopts a “one-size-fits-all” approach, recent studies have shown the potential of per-instance algorithm configuration (PIAC) methods in these settings. In this work, we evaluate and compare the performance of different PIAC methods in this context, namely Hydra—a state-of-the-art PIAC method—and a simpler case-based reasoning (CBR) approach. We evaluate the impact of the tuning time budget and/or the number of unique problem instances used for training on each of the method’s performance and robustness. Our experiments show that whilst Hydra fails to improve upon the default algorithm configuration, the CBR method can lead to 16% performance increase using as few as 100 training instances. Following these findings, we evaluate Hydra’s methodology when applied to homogenous instance spaces. This analysis shows the limitations of Hydra’s inference mechanisms in these settings and showcases the advantages of distance-based approaches used in CBR.

Reinforcement learning-trained optimisers and Bayesian optimisation for online particle accelerator tuning

Online tuning of particle accelerators is a complex optimisation problem that continues to require manual intervention by experienced human operators. Autonomous tuning is a rapidly expanding field of research, where learning-based methods like Bayesian optimisation (BO) hold great promise in improving plant performance and reducing tuning times. At the same time, reinforcement learning (RL) is a capable method of learning intelligent controllers, and recent work shows that RL can also be used to train domain-specialised optimisers in so-called reinforcement learning-trained optimisation (RLO). In parallel efforts, both algorithms have found successful adoption in particle accelerator tuning. Here we present a comparative case study, assessing the performance of both algorithms while providing a nuanced analysis of the merits and the practical challenges involved in deploying them to real-world facilities. Our results will help practitioners choose a suitable learning-based tuning algorithm for their tuning tasks, accelerating the adoption of autonomous tuning algorithms, ultimately improving the availability of particle accelerators and pushing their operational limits.

Boosted multilayer feedforward neural network with multiple output layers

This research introduces the Boosted Ensemble deep Multi-Layer Layer Perceptron (EdMLP) architecture with multiple output layers, a novel enhancement for the traditional Multi-Layer Perceptron (MLP). By adopting a layer-wise training approach, EdMLP enables the integration of boosting techniques within a single model, treating each layer as a weak learner, resulting in substantial performance gains. Additionally, the inclusion of layer-wise hyperparameter tuning allows optimization of individual layers thereby reducing the tuning time. Furthermore, the ensemble deep architecture’s versatility can be extended to other neural network-based models, such as the Self Normalized Network (SNN) where experiments demonstrate substantial performance enhancements yielded by the EdSNN compared to the standard original SNN model. This research underscores the potential of the EdMLP, and the Ed architecture in general as a powerful tool for improving the performance of various multilayer feedforward neural network models. The source code of this work is publicly accessible from the authors GitHub.

DEVELOPMENT AND ASSESSMENT OF MANUAL AND AUTOMATED PID CONTROLLERS FOR THE OPTIMUM PRODUCTION OF ETHYLENE GLYCOL IN A CSTR

Industrially, one of the techniques used to enhance the optimum production of petrochemicals is the configuration of process control units such as gas chromatography (G.C. Analyzer) for concentration and thermocouple for temperature in the production system. This research focused on the application of the principles of mass and energy conservation in the development of dynamic or unsteady state models for predicting the behaviour of the system or process involved in the production of ethylene glycol from a reaction involving oxidation of ethylene-to-ethylene oxide which further undergoes hydrolysis to produce ethylene glycol in a continuous stirred tank reactor (CSTR). The high mathematical complexities of the mass and energy balance models of the Process as a result of the reaction kinetic scheme of the non-elementary reaction of the process which integrated both linear and non-linear terms into the models were resolved by the application of the principles of Taylor’s series expansion, Laplace Transform, partial fraction and matrix in the development of the dynamic models. Process simulation tool (Simulink) was utilized in the configuration of closed-loop systems with thermocouple and G. C. Analyzer, at the feed point of the reactor for process variables (temperature and concentration) control during the process using proportional, integral and derivative (PID) as controller parameters. The results of the process behaviour showed that manual tuning of different controller parameters experienced a high level of fluctuations and would not attain its stability or steady state even at a processing time above 16 seconds. Whereas, automatic tuning of desired controller gains of 1.361, 1.056 and 0.4109 for K P, K I and K D for temperature and pressure with a tuning time of 143.9 Seconds, requires a maximum time of 8 seconds for the system to attain stability. This study showed that automated tuning of the controller resulted in better performance characteristics for optimum production of ethylene glycol in a non-isotheral continuous stirred tank reactor within the shortest possible time.

Learning Advanced Locomotion for Quadrupedal Robots: A Distributed Multi-Agent Reinforcement Learning Framework with Riemannian Motion Policies

Recent advancements in quadrupedal robotics have explored the motor potential of these machines beyond simple walking, enabling highly dynamic skills such as jumping, backflips, and even bipedal locomotion. While reinforcement learning has demonstrated excellent performance in this domain, it often relies on complex reward function tuning and prolonged training times, and the interpretability is not satisfactory. Riemannian motion policies, a reactive control method, excel in handling highly dynamic systems but are generally limited to fully actuated systems, making their application to underactuated quadrupedal robots challenging. To address these limitations, we propose a novel framework that treats each leg of a quadrupedal robot as an intelligent agent and employs multi-agent reinforcement learning to coordinate the motion of all four legs. This decomposition satisfies the conditions for utilizing Riemannian motion policies and eliminates the need for complex reward functions, simplifying the learning process for high-level motion modalities. Our simulation experiments demonstrate that the proposed method enables quadrupedal robots to learn stable locomotion using three, two, or even a single leg, offering advantages in training speed, success rate, and stability compared to traditional approaches, and better interpretability. This research explores the possibility of developing more efficient and adaptable control policies for quadrupedal robots.

Enhancing genome-wide populus trait prediction through deep convolutional neural networks.

As a promising model, genome-based plant breeding has greatly promoted the improvement of agronomic traits. Traditional methods typically adopt linear regression models with clear assumptions, neither obtaining the linkage between phenotype and genotype nor providing good ideas for modification. Nonlinear models are well characterized in capturing complex nonadditive effects, filling this gap under traditional methods. Taking populus as the research object, this paper constructs a deep learning method, DCNGP, which can effectively predict the traits including 65 phenotypes. The method was trained on three datasets, and compared with other four classic models-Bayesian ridge regression (BRR), Elastic Net, support vector regression, and dualCNN. The results show that DCNGP has five typical advantages in performance: strong prediction ability on multiple experimental datasets; the incorporation of batch normalization layers and Early-Stopping technology enhancing the generalization capabilities and prediction stability on test data; learning potent features from the data and thus circumventing the tedious steps of manual production; the introduction of a Gaussian Noise layer enhancing predictive capabilities in the case of inherent uncertainties or perturbations; fewer hyperparameters aiding to reduce tuning time across datasets and improve auto-search efficiency. In this way, DCNGP shows powerful predictive ability from genotype to phenotype, which provide an important theoretical reference for building more robust populus breeding programs.

Study of Air Ordering Procedure for Information Broadcasting

Wireless data broadcasting is a new way to share public information with many mobile users. To measure how well these systems work, we look at two main things: how long it takes to access the information (access latency) and how long the device has to stay active (tuning time). Indexing technology helps reduce tuning time by letting devices find information quickly and then switch to a low-power mode while waiting. Different indexing methods work better or worse, so comparing them can be hard. In this paper, I look at and compare some popular indexing methods for wireless data broadcasting, such as distributed index, exponential index, hash table, and Huffman tree index. I use probability theory to analyze how each method performs. DOI: https://doi.org/10.52783/tjjpt.v45.i02.6393

Local Bayesian optimization for controller tuning with crash constraints

Abstract Controller tuning is crucial for closed-loop performance but often involves manual adjustments. Although Bayesian optimization (BO) has been established as a data-efficient method for automated tuning, applying it to large and high-dimensional search spaces remains challenging. We extend a recently proposed local variant of BO to include crash constraints, where the controller can only be successfully evaluated in an a-priori unknown feasible region. We demonstrate the efficiency of the proposed method through simulations and hardware experiments. Our findings showcase the potential of local BO to enhance controller performance and reduce the time and resources necessary for tuning.

Coupling speckle noise suppression with image classification for deep-learning-aided ultrasound diagnosis

Objective. During deep-learning-aided (DL-aided) ultrasound (US) diagnosis, US image classification is a foundational task. Due to the existence of serious speckle noise in US images, the performance of DL models may be degraded. Pre-denoising US images before their use in DL models is usually a logical choice. However, our investigation suggests that pre-speckle-denoising is not consistently advantageous. Furthermore, due to the decoupling of speckle denoising from the subsequent DL classification, investing intensive time in parameter tuning is inevitable to attain the optimal denoising parameters for various datasets and DL models. Pre-denoising will also add extra complexity to the classification task and make it no longer end-to-end. Approach. In this work, we propose a multi-scale high-frequency-based feature augmentation (MSHFFA) module that couples feature augmentation and speckle noise suppression with specific DL models, preserving an end-to-end fashion. In MSHFFA, the input US image is first decomposed to multi-scale low-frequency and high-frequency components (LFC and HFC) with discrete wavelet transform. Then, multi-scale augmentation maps are obtained by computing the correlation between LFC and HFC. Last, the original DL model features are augmented with multi-scale augmentation maps. Main results. On two public US datasets, all six renowned DL models exhibited enhanced F1-scores compared with their original versions (by 1.31%–8.17% on the POCUS dataset and 0.46%–3.89% on the BLU dataset) after using the MSHFFA module, with only approximately 1% increase in model parameter count. Significance. The proposed MSHFFA has broad applicability and commendable efficiency and thus can be used to enhance the performance of DL-aided US diagnosis. The codes are available at https://github.com/ResonWang/MSHFFA.

Continuous adiabatic frequency conversion for FMCW-LiDAR

Continuous tuning of the frequency of laser light serves as the fundamental basis for a myriad of applications spanning basic scientific research to industrial settings. These applications encompass endeavors such as the detection of gravitational waves, the development of precise optical clocks, environmental monitoring for health and ecological purposes, as well as distance measurement techniques. However, achieving a broad tuning range exceeding 100 GHz along with sub-microsecond tuning times, inherent linearity in tuning, and coherence lengths beyond 10 m presents significant challenges. Here, we demonstrate that electro-optically driven adiabatic frequency converters utilizing high-Q microresonators fabricated from lithium niobate possess the capability to convert arbitrary voltage signals into frequency chirps with temporal resolutions below 1 µs. The temporal evolution of the frequency correlates accurately with the applied voltage signal. We have achieved to generate 200-ns-long frequency chirps with deviations of less than 1 % from perfect linearity without requiring supplementary measures. The coefficient of determination is R2>0.999\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^2>0.999$$\\end{document}. Moreover, the coherence length of the emitted light exceeds 20 m. To validate these findings, we employ the linear frequency sweeps for Frequency-Modulated Continuous Wave (FMCW) LiDAR covering distances ranging from 0.5 to 10 m. Leveraging the demonstrated nanosecond-level tuning capabilities, coupled with the potential to tune the eigenfrequency of lithium-niobate-based resonators by several hundred GHz, our results show that electro-optically driven adiabatic frequency converters can be used in applications that require ultrafast and flexible continuous frequency tuning characterized by inherent linearity and substantial coherence length.

Longitudinal vibration estimation of a mine hoist using a hybrid signal fusion method combining UKF, ND and improved DE

The monitoring of cage longitudinal vibration can directly indicate the operational status of mine hoists. However, it is always challenging to collect the sensor signals of moving cages with high dynamic characteristics in real time from complex working environments using traditional monitoring methods. In this study, a more practical hybrid signal fusion approach is proposed to realize estimation of cage longitudinal vibration from a low sampling rate acceleration acquisition signal and a low cost encoder signal for state estimation. A nonlinear differentiator is applied to extract encoder differential signals and expand observation variables. An unscented Kalman observer based on nonlinear mine hoist model is designed to estimate the unknown state. To overcome the influence of the uncertain parameters, an improved differential evolution (DE) algorithm combining parameter adaptive method, reverse learning competition scheme and multiple parallel populations strategy is proposed to find unknown parameters of the observation model and autotune the parameters of the algorithms by using low sampling rate acceleration. Sensor data of the simulated experiment platform were collected and processed by the xPC system to validate the effectiveness of the proposed strategy. The experimental results showed that the improved DE (IDE) algorithm had a faster mean time for parameter tuning and the smallest fitness value compared to the standard DE, the particle swarm optimization algorithm and the genetic algorithm. Moreover, the longitudinal vibration estimation system, after parameter tuning by the IDE optimization algorithm, could achieve the purpose of signal estimation, with a smaller estimation error and a better estimation effect.

A novel ensemble model of multi-class credit assessment based on multi-source fusion theory

With the development of artificial intelligence technology, the assessment method based on machine learning, especially the ensemble learning method, has attracted more and more attention in the field of credit assessment. However, most of the ensemble assessment models are complex in structure and costly in time for parameter tuning, few of them break through the limitations of lightweight, universal and efficient. This paper present a new ensemble model for personal credit assessment. First, considering the conflicts and differences among multiple sources of information, a new method is proposed to correct the category prior information by using the difference measure. Then, the revised prior information is fused with the current sample information with the help of Bayesian data fusion theory. The model can integrate the advantages of multiple benchmark classifiers to reduce the interference of uncertain information. To verify the effectiveness of the proposed model, several typical ensemble classification models are selected and empirically studied using real customer credit data from a commercial bank in China, and the results show that among various assessment criteria: the proposed model not only effectively improves the multi-class classification performance, but also outperforms other advanced multi-class classification credit assessment models in terms of parameter tuning and generalizability. This paper supports commercial banks and other financial institutions examination and approval work.

Cross-Feature Transfer Learning for Efficient Tensor Program Generation

Tuning tensor program generation involves navigating a vast search space to find optimal program transformations and measurements for a program on the target hardware. The complexity of this process is further amplified by the exponential combinations of transformations, especially in heterogeneous environments. This research addresses these challenges by introducing a novel approach that learns the joint neural network and hardware features space, facilitating knowledge transfer to new, unseen target hardware. A comprehensive analysis is conducted on the existing state-of-the-art dataset, TenSet, including a thorough examination of test split strategies and the proposal of methodologies for dataset pruning. Leveraging an attention-inspired technique, we tailor the tuning of tensor programs to embed both neural network and hardware-specific features. Notably, our approach substantially reduces the dataset size by up to 53% compared to the baseline without compromising Pairwise Comparison Accuracy (PCA). Furthermore, our proposed methodology demonstrates competitive or improved mean inference times with only 25–40% of the baseline tuning time across various networks and target hardware. The attention-based tuner can effectively utilize schedules learned from previous hardware program measurements to optimize tensor program tuning on previously unseen hardware, achieving a top-5 accuracy exceeding 90%. This research introduces a significant advancement in autotuning tensor program generation, addressing the complexities associated with heterogeneous environments and showcasing promising results regarding efficiency and accuracy.