Stochastic Machine Research Articles

Stochastic machine learning via sigma profiles to build a digital chemical space

This work establishes a different paradigm on digital molecular spaces and their efficient navigation by exploiting sigma profiles. To do so, the remarkable capability of Gaussian processes (GPs), a type of stochastic machine learning model, to correlate and predict physicochemical properties from sigma profiles is demonstrated, outperforming state-of-the-art neural networks previously published. The amount of chemical information encoded in sigma profiles eases the learning burden of machine learning models, permitting the training of GPs on small datasets which, due to their negligible computational cost and ease of implementation, are ideal models to be combined with optimization tools such as gradient search or Bayesian optimization (BO). Gradient search is used to efficiently navigate the sigma profile digital space, quickly converging to local extrema of target physicochemical properties. While this requires the availability of pretrained GP models on existing datasets, such limitations are eliminated with the implementation of BO, which can find global extrema with a limited number of iterations. A remarkable example of this is that of BO toward boiling temperature optimization. Holding no knowledge of chemistry except for the sigma profile and boiling temperature of carbon monoxide (the worst possible initial guess), BO finds the global maximum of the available boiling temperature dataset (over 1,000 molecules encompassing more than 40 families of organic and inorganic compounds) in just 15 iterations (i.e., 15 property measurements), cementing sigma profiles as a powerful digital chemical space for molecular optimization and discovery, particularly when little to no experimental data is initially available.

Stochastic Machine Learning Models for Mutation Rate Analysis of Malignant Cancer Cells in Patients with Acute Lymphoblastic Leukemia

Stochastic Machine Learning Models for Mutation Rate Analysis of Malignant Cancer Cells in Patients with Acute Lymphoblastic Leukemia

Machine learning attack detection based-on stochastic classifier methods for enhancing of routing security in wireless sensor networks

Wireless Sensor Networks (WSNs) are vulnerable to attacks during data transmission, and many techniques have been proposed to detect and secure routing data. In this paper, we introduce a novel stochastic predictive machine learning approach designed to discern untrustworthy events and unreliable routing attributes, aiming to establish an artificial intelligence-based attack detection system for WSNs. Our methodology leverages real-time analysis of the features of simulated WSN routing data. By integrating Hidden Markov Models (HMM) with Gaussian Mixture Models (GMM), we develop a robust classification framework. This framework effectively identifies outliers, pinpoints malicious network behaviors from their origins, and categorizes them as either trusted or untrusted network activities. In addition, dimensionality reduction techniques are used to improve interpretability, reduce computation and processing time, extract uncorrelated features from network data, and optimize performances. The main advantage of our approach is to establish an efficient stochastic machine learning method capable of analyzing and filtering WSN traffic to prevent suspicious and unsafe data, reduce the large dissimilarity in the collected routing features, and rapidly detect attacks before they occur. In this work, we exploit a well-tuned data set that provides a lot of routing information without losing any data. The experimental results show that the proposed stochastic attack detection system can effectively identify and categorize anomalies in wireless sensor networks with high accuracy. The classification rates of the system were found to be around 83.65%, 84.94% and 94.55%, which is significantly better than the existing classification approaches. Furthermore, the proposed system showed a positive prediction value of 11.84% higher than the existing approaches.

Flexible job shop scheduling with stochastic machine breakdowns by an improved tuna swarm optimization algorithm

In job-shop production environments, machine breakdowns are a significant factor in reducing productivity. Existing approaches seldom consider algorithm improvement and rescheduling scheme design in an integrated manner, and lack stability considerations. This paper addresses the flexible job shop scheduling problem with random machine breakdowns, aiming to produce a stable rescheduling scheme that minimizes a combined index of maximum completion time and stability. The paper innovatively applies the tuna swarm optimization algorithm to the flexible job shop scheduling problem, proposing an efficient and superior improved version called the genetic chaos levy nonlinear tuna swarm optimization (GCLNTSO) algorithm. Three stability metrics are designed to guide the generation of efficient and stable rescheduling schemes. A rescheduling scheme is proposed that combines right-shift rescheduling with complete rescheduling. The proposed scheme is benchmarked against Brandymalter and Kacem’s problems, and compared with other algorithms from the literature. The results demonstrate that the GCLNTSO algorithm outperforms other algorithms in terms of both performance and stability.

Solving the stochastic machine assignment problem with a probability-based objective: problem formulation, solution method and practical applications

Solving the stochastic machine assignment problem with a probability-based objective: problem formulation, solution method and practical applications

Solving the Stochastic Machine Assignment Problem with Probability-based Objectives: Problem Formulation, Solution Method, and Practical Applications

Solving the Stochastic Machine Assignment Problem with Probability-based Objectives: Problem Formulation, Solution Method, and Practical Applications

Stochastic Machine Learning Based Attacks Detection System in Wireless Sensor Networks

Stochastic Machine Learning Based Attacks Detection System in Wireless Sensor Networks

Robust job shop scheduling with machine unavailability due to random breakdowns and condition-based maintenance

This paper presents a novel solution approach for a variant of the job shop scheduling problem with machine unavailability due to both condition-based preventive maintenance and corrective maintenance following random breakdowns. We first provide an exact mathematical formulation of the problem under simplifying assumptions, namely that the number of breakdowns for each job position on each machine is known, the degradation rates are fixed, and the preventive and corrective maintenance durations are deterministic parameters. Moreover, to handle the more realistic case of stochastic machine degradation, random breakdowns, and uncertain maintenance durations, a simulation-optimisation algorithm is proposed. The real makespan function is first approximated using multiple surrogate measures, which are optimised through independent genetic algorithms. Then, the fittest solutions obtained from these surrogate measures are simulated, and the best among them is added to an elite list, which is included in the genetic algorithms' populations for the next iteration. Schedule robustness is ensured by using an objective function that consists of the weighted average of the expected makespan and its 90th percentile. Furthermore, to reduce the likelihood of falling into a local optimum, a stopping criterion based on simulated annealing is implemented. Numerical experimentation on extended benchmark instances confirmed the validity of the mathematical formulation and the favourable performance of the proposed simulation-optimisation algorithm in terms of computational time and solution quality.

Calibrated forecasts of quasi‐periodic climate processes with deep echo state networks and penalized quantile regression

AbstractAmong the most relevant processes in the Earth system for human habitability are quasi‐periodic, ocean‐driven multi‐year events whose dynamics are currently incompletely characterized by physical models, and hence poorly predictable. This work aims at showing how (1) data‐driven, stochastic machine learning approaches provide an affordable yet flexible means to forecast these processes; (2) the associated uncertainty can be properly calibrated with fast ensemble‐based approaches. While the methodology introduced and discussed in this work pertains to synoptic scale events, the principle of augmenting incomplete or highly sensitive physical systems with data‐driven models to improve predictability is far more general and can be extended to environmental problems of any scale in time or space.

Minimizing the expected maximum lateness for a job shop subject to stochastic machine breakdowns

This paper addresses a stochastic job shop scheduling problem with sequence-dependent setup times, aiming to minimize the expected maximum lateness. The stochastic nature is modeled by considering uncertain times between failures (TBF) and uncertain times to repair (TTR). To tackle this problem, a simheuristic approach is proposed, which combines a tabu search (TS) algorithm with Monte Carlo simulation. A total of 320 instances were used to conduct multiple experiments. Instances were generated with two distributions to study the behavior of stochastic TTR and TBF under log-normal and exponential distributions. Firstly, the performance of the simheuristic was evaluated for small instances by comparing it with the simulation of optimal solutions obtained with a mixed-integer linear programming (MILP) model. The simheuristic approach demonstrated an average improvement of around 7% compared to the simulation of MILP model solutions. Secondly, the simheuristic performance was evaluated for medium and large-size instances by comparing it with the simulation of the solutions obtained by the earliest due date (EDD) and process time plus work in the next queue plus negative slack (PT + WINQ + SL) dispatching rules. The results showed an average improvement of around 11% compared to EDD and 14% compared to PT + WINQ + SL. Furthermore, the results highlight that even when the two distributions have the same expected value and coefficient of variation, they can yield different expected maximum lateness values. This emphasizes the importance of precise distribution fitting when solving real cases to achieve effective scheduling performance.

Impact study of numerical discretization accuracy on parameter reconstructions and model parameter distributions

In optical nano metrology numerical models are used widely for parameter reconstructions. Using the Bayesian target vector optimization method we fit a finite element numerical model to a Grazing Incidence x-ray fluorescence data set in order to obtain the geometrical parameters of a nano structured line grating. Gaussian process, stochastic machine learning surrogate models, were trained during the reconstruction and afterwards sampled with a Markov chain Monte Carlo sampler to determine the distribution of the reconstructed model parameters. The numerical discretization parameters of the used finite element model impact the numerical discretization error of the forward model. We investigated the impact of the polynomial order of the finite element ansatz functions on the reconstructed parameters as well as on the model parameter distributions. We showed that such a convergence study allows to determine numerical parameters which allows for efficient and accurate reconstruction results.

SMLFire1.0: a stochastic machine learning (SML) model for wildfire activity in the western United States

Abstract. The annual area burned due to wildfires in the western United States (WUS) increased by more than 300 % between 1984 and 2020. However, accounting for the nonlinear, spatially heterogeneous interactions between climate, vegetation, and human predictors driving the trends in fire frequency and sizes at different spatial scales remains a challenging problem for statistical fire models. Here we introduce a novel stochastic machine learning (SML) framework, SMLFire1.0, to model observed fire frequencies and sizes in 12 km × 12 km grid cells across the WUS. This framework is implemented using mixture density networks trained on a wide suite of input predictors. The modeled WUS fire frequency matches observations at both monthly (r=0.94) and annual (r=0.85) timescales, as do the monthly (r=0.90) and annual (r=0.88) area burned. Moreover, the modeled annual time series of both fire variables exhibit strong correlations (r≥0.6) with observations in 16 out of 18 ecoregions. Our ML model captures the interannual variability and the distinct multidecade increases in annual area burned for both forested and non-forested ecoregions. Evaluating predictor importance with Shapley additive explanations, we find that fire-month vapor pressure deficit (VPD) is the dominant driver of fire frequencies and sizes across the WUS, followed by 1000 h dead fuel moisture (FM1000), total monthly precipitation (Prec), mean daily maximum temperature (Tmax), and fraction of grassland cover in a grid cell. Our findings serve as a promising use case of ML techniques for wildfire prediction in particular and extreme event modeling more broadly. They also highlight the power of ML-driven parameterizations for potential implementation in fire modules of dynamic global vegetation models (DGVMs) and earth system models (ESMs).

Stochastic modeling of physical drag coefficient – Its impact on orbit prediction and space traffic management

Ambitious satellite constellation projects by commercial entities and the ease of access to space in recent times have led to a dramatic proliferation of low-Earth space traffic. It jeopardizes space safety and long-term sustainability, necessitating better space domain awareness (SDA). Correct modeling of uncertainties in force models and orbital states, among other things, is an essential part of SDA. For objects in the low-Earth orbit (LEO) region, the uncertainty in the orbital dynamics mainly emanate from limited knowledge of the atmospheric drag-related parameters and variables. In this paper, which extends the work by Paul et al. (2021), we develop a feed-forward deep neural network model for the prediction of the satellite drag coefficient for the full range of satellite attitude (i.e., satellite pitch ∈ (-90°,+90°) and satellite yaw ∈ (0°,+360°)). The model simultaneously predicts the mean and the standard deviation and is well-calibrated. We use numerically simulated physical drag coefficient data for training our neural network. The numerical simulations are carried out using the test particle Monte Carlo method using the diffuse reflection with incomplete accommodation gas-surface interaction model. Modeling is carried out for the well-known CHAllenging Minisatellite Payload (CHAMP) satellite. Finally, we use the Monte Carlo approach to propagate CHAMP over a three-day period under various modeling scenarios to investigate the distribution of radial, along-track, and cross-track orbital errors caused by drag coefficient uncertainty. The key takeaways of this paper are - (a) a constant drag coefficient cannot be used for reliable SDA purposes, and (b) stochastic machine learning models allow for the computation of drag coefficients in a timely manner while providing reliable uncertainty estimates.

Equivalence between finite state stochastic machine, non-dissipative and dissipative tight-binding and Schrödinger model

The mathematical equivalence between finite state stochastic machine and non-dissipative and dissipative quantum tight-binding and Schrödinger model is derived. Stochastic Finite state machine is also expressed by classical epidemic model and can reproduce the quantum entanglement emerging in the case of electrostatically coupled qubits described by von-Neumann entropy both in non-dissipative and dissipative case. The obtained results show that quantum mechanical phenomena might be simulated by classical statistical model as represented by finite state stochastic machine. It includes the quantum like entanglement and superposition of states. Therefore coupled epidemic models expressed by classical systems in terms of classical physics can be the base for possible incorporation of quantum technologies and in particular for quantum like computation and quantum like communication. The classical density matrix is derived and described by the equation of motion in terms of anticommutator. Existence of Rabi like oscillations is pointed in classical epidemic model. Furthermore, the existence of Aharonov–Bohm effect in quantum systems (Becker et al., 2019) can also be reproduced by the classical epidemic model or in broader sense by finite state stochastic machine. Every quantum system made from quantum dots and described by simplistic tight-binding model by use of position-based qubits can be effectively described by classical statistical model encoded in finite stochastic state machine with very specific structure of S matrix that has twice bigger size as it is the case of quantum matrix Hamiltonian. Furthermore, the description of linear and non-linear stochastic finite state machine is mapped to tight-binding and Schrödinger model. The concept of N dimensional complex time is incorporated into tight-binding model, so the description of dissipation in most general case is possible. Obtained results help in approximation of non-dissipative or dissipative quantum mechanical phenomena by classical statistical physics expressed by finite state stochastic machine. Furthermore, the equivalence of Wannier tight-binding dissipative or non-dissipative formalism to dissipative or non-dissipative Schrödinger formalism and to finite state stochastic machine was shown.

Evaluation of the Robustness for Integrated Production Scheduling and Maintenance Planning Problem

In this paper, we consider the surrogate measures of robustness for the integration of production scheduling and preventive maintenance planning problem with stochastic machine failures. First, the influence of two categories of maintenance on operation completion time is analyzed. Since it does not come with an exact solution, a novel measure algorithm for evaluating two types of robustness simultaneously is proposed, based on the internal relationships among the scheduling structure, the number and sequence of preventive maintenance activities, the probability and downtime of failures, and the expected completion time of operation. Extensive experiments are conducted on 19 benchmark problems with random machine breakdowns. Experimental results first show that the correlation between our algorithm and Monte Carlo simulation in two robustness indicators is above 99%, and the former computes much faster than the latter. A thorough comparison is made with the other three surrogate robustness measures, which further proves the accuracy of our algorithm. Additional experiments also confirm the benefits by integrating preventive maintenance.

Sustainable production model with stochastic machine breakdown using smart manufacturing under circular economy

<abstract> <p>Machine breakdown usually implies unexpected physical damage to machinery due to any reason which requires fixing or replacement to continue the process. This article presentsan investigation of a sustainable model with stochastic machine breakdown. To reduce the risk of disruption,a smart manufacturing system is used. Considering the environmental issues faced by people,the model is developed under a circular economy through end-of-life treatment to recapture the value of the product,labor and resources. The used buyback products are collected,out of which items in good condition are remanufactured and sold in another market while the rest are salvaged. As the production process is not perfectly reliable,the serviceable products go through an automated inspection process,and imperfect items are reworked. A mathematical model is developed for deterioration of items to analyze the optimal replenishment policies,and the results have been illustrated with numerical verification. Based on the analysis,some managerial insights have been provided for decision-makers.</p> </abstract>

Robust-stable scheduling in dynamic flow shops based on deep reinforcement learning

AbstractThis proof-of-concept study provides a novel method for robust-stable scheduling in dynamic flow shops based on deep reinforcement learning (DRL) implemented with OpenAI frameworks. In realistic manufacturing environments, dynamic events endanger baseline schedules, which can require a cost intensive re-scheduling. Extensive research has been done on methods for generating proactive baseline schedules to absorb uncertainties in advance and in balancing the competing metrics of robustness and stability. Recent studies presented exact methods and heuristics based on Monte Carlo experiments (MCE), both of which are very computationally intensive. Furthermore, approaches based on surrogate measures were proposed, which do not explicitly consider uncertainties and robustness metrics. Surprisingly, DRL has not yet been scientifically investigated for generating robust-stable schedules in the proactive stage of production planning. The contribution of this article is a proposal on how DRL can be applied to manipulate operation slack times by stretching or compressing plan durations. The method is demonstrated using different flow shop instances with uncertain processing times, stochastic machine failures and uncertain repair times. Through a computational study, we found that DRL agents achieve about 98% result quality but only take about 2% of the time compared to traditional metaheuristics. This is a promising advantage for the use in real-time environments and supports the idea of improving proactive scheduling methods with machine learning based techniques.

Integrated Scheduling of the Production and Maintenance of Parallel Machine Job-shop Considering Stochastic Machine Breakdowns

The integrated scheduling of production and maintenance can make equipment maintenance in line with the production pace, so as to effectively prevent anormal interruptions of the production process due to equipment failure, and ensure the smooth implementation of the production scheduling plan. Aiming at the parallel machine job-shop environment, and considering stochastic machine failures and different degradation speeds of parallel machines, this paper introduced the minimal maintenance and preventive maintenance strategies to establish an integrated scheduling model for production and maintenance, designed a genetic algorithm based on process coding and binary hybrid coding to solve the model, and verified the correctness of the proposed model and the effectiveness of the algorithm through an instance. This study provided an effective decision-making method for parallel machine job-shop scheduling problems.

Single-cell analysis of habituation in Stentor coeruleus

Although learning is often viewed as a unique feature of organisms with complex nervous systems, single-celled organisms also demonstrate basic forms of learning. The giant ciliate Stentor coeruleus responds to mechanical stimuli by contracting into a compact shape, presumably as a defense mechanism. When a Stentor cell is repeatedly stimulated at a constant level of force, it will learn to ignore that stimulus but will still respond to stronger stimuli. Prior studies of habituation in Stentor reported a graded response, suggesting that cells transition through a continuous range of response probabilities. By analyzing single cells using an automated apparatus to deliver calibrated stimuli, we find that habituation occurs via a single step-like switch in contraction probability within each cell, with the graded response in a population arising from the random distribution of switching times in individual cells. This step-like response allows Stentor behavior to be represented by a simple two-state model whose parameters can be estimated from experimental measurements. We find that transition rates depend on stimulus force and also on the time between stimuli. The ability to measure the behavior of the same cell to the same stimulus allowed us to quantify the functional heterogeneity among single cells. Together, our results suggest that the behavior of Stentor is governed by a two-state stochastic machine whose transition rates are sensitive to the time series properties of the input stimuli.

Wind speed and global radiation forecasting based on differential, deep and stochastic machine learning of patterns in 2-level historical meteo-quantity sets

AbstractAccurate forecasting of wind speed and solar radiation can help operators of wind farms and Photo-Voltaic (PV) plants prepare efficient and practicable production plans to balance the supply with demand in the generation and consumption of Renewable Energy (RE). Reliable Artificial Intelligence (AI) forecast models can minimize the effect of wind and solar power fluctuations, eliminating their intermittent character in system dispatch planning and utilization. Intelligent wind and solar energy management is essential in load scheduling and decision-making processes to meet user requirements. The proposed 24-h prediction schemes involve the beginning detection and secondary similarity re-evaluation of optimal day-data sequences, which is a notable incremental improvement against state-of-the-art in the consequent application of statistical AI learning. 2-level altitude measurements allow the identification of data relationships between two surface layers (hill and lowland) and adequate interpretation of various meteorological situations, whose differentiate information is used by AI models to recognize upcoming changes in the mid-term day horizon. Observations at two professional meteorological stations comprise specific quantities, of which the most valuable are automatically selected as input for the day model. Differential learning is a novel designed unconventional neurocomputing approach that combines derivative components produced in selected network nodes in the weighted modular output. The complexity of the node-stepwise composed model corresponds to the patterns included in the training data. It allows for representation of high uncertain and nonlinear dynamic systems, dependent on local RE production, not substantially reducing the input vector dimensionality leading to model over simplifications as standard AI does. Available angular and frequency time data (e.g., wind direction, humidity, and irradiation cycles) are combined with the amplitudes to solve reduced Partial Differential Equations (PDEs), defined in network nodes, in the periodical complex form. This is a substantial improvement over the previous publication design. The comparative results show better efficiency and reliability of differential learning in representing the modular uncertainty and PDE dynamics of patterns on a day horizon, taking into account recent deep and stochastic learning. A free available C++ parametric software together with the processed meteo-data sets allow additional comparisons with the presented model results.