Parallel Inference for Real-Time Machine Learning Applications

Hyperparameter tuning is a key step in developing high-performing machine learning models, but searching large hyperparameter spaces requires extensive computation using standard sequential methods. This work analyzes the performance gains from parallel versus sequential hyperparameter optimization. Using scikit-learn’s Randomized SearchCV, this project tuned a Random Forest classifier for fake news detection via randomized grid search. Setting n_jobs to -1 enabled full parallelization across CPU cores. Results show the parallel implementation achieved over 5× faster CPU times and 3× faster total run times compared to sequential tuning. However, test accuracy slightly dropped from 99.26% sequentially to 99.15% with parallelism, indicating a trade-off between evaluation efficiency and model performance. Still, the significant computational gains allow more extensive hyperparameter exploration within reasonable timeframes, outweighing the small accuracy decrease. Further analysis could better quantify this trade-off across different models, tuning techniques, tasks, and hardware.

ORDER-PICKING OPTIMIZATION FOR AUTOMATED PICKING SYSTEM WITH PARALLEL DISPENSERS

Based on the characteristics of parallel dispensers in automated picking system, an order-picking optimization problem is presented. Firstly, the working principle of parallel dispensers is introduced, which implies the time cost of picking each order is influenced by the order-picking sequence. So the order-picking optimization problem can be classified as a dynamic traveling salesman problem （TSP）. Then a mathematical model of the problem is established and an improved max-min ant system （MMAS） is adopted to solve the model. The improvement includes two aspects. One is that the initial assignment of ants depends on a probabilistic formula instead of a random deployment; the other is that the heuristic factor is expressed by the extra picking time of each order instead of the total. At last, an actual simulation is made on an automated picking system with parallel dispensers. The simulation results proved the optimization value and the validity of improvement on MMAS.

Experimental Characterization of Operation of a Waist-Trunk System with Parallel Manipulators

An extensive research activity has been focused on the upper and lower limbs of humanoid robots. However, due to mechanical design difficulties and complex control of multi-body system, the torso of humanoid robot is somehow a neglected or simplified design part. In this paper, operation performance of a new waist-trunk system as torso for humanoid robots is presented through results of lab experimental tests. The proposed waist-trunk system is composed of two 3 DOFs （degrees of freedom） parallel manipulators, which are connected in a serial chain architecture. A prototype is built by using two prototypes of CaPaMan （Cassino Parallel Manipulator）, which are convenient stiff architectures with easy-operation characteristics. Experimental tests are carried out with the aims to imitate lateral-bending and transverse-rotation movements of human torso. Operation performances like displacements, accelerations, and actuation torque are measured for a performance evaluation and design characterization of the used manipulator solution imitating human torso. Experimental test results are illustrated and discussed to show the practical operation feasibility of the proposed architecture and the operation characteristics of the built prototype.

Parallel Reinforcement Learning-Based Energy Efficiency Improvement for a Cyber-Physical System

As a complex and critical cyber-physical system(CPS),the hybrid electric powertrain is significant to mitigate air pollution and improve fuel economy.Energy management strategy(EMS)is playing a key role to improve the energy efficiency of this CPS.This paper presents a novel bidirectional long shortterm memory(LSTM)network based parallel reinforcement learning(PRL)approach to construct EMS for a hybrid tracked vehicle(HTV).This method contains two levels.The high-level establishes a parallel system first,which includes a real powertrain system and an artificial system.Then,the synthesized data from this parallel system is trained by a bidirectional LSTM network.The lower-level determines the optimal EMS using the trained action state function in the model-free reinforcement learning(RL)framework.PRL is a fully data-driven and learning-enabled approach that does not depend on any prediction and predefined rules.Finally,real vehicle testing is implemented and relevant experiment data is collected and calibrated.Experimental results validate that the proposed EMS can achieve considerable energy efficiency improvement by comparing with the conventional RL approach and deep RL.

Selective maintenance problem for series–parallel system under economic dependence

In view of the high complexity of the objective world, an economic dependence between subsystems(paired and unpaired) is proposed, and then the maintenance cost and time under different economic dependences are formulated in a simple and consistent manner. Selective maintenance problem under economic dependence(EDSMP) is presented based on a series–parallel system in this paper. A case study shows that the system reliability is promoted to a certain extent, which can validate the validity of the EDSMP model. The influence of the ratio of set-up cost on system performance is mainly discussed under different economic dependences. Several existing improvements of classical exhaust algorithm are further modified to solve a large sized EDSMP rapidly. Experimental results illustrate that these improvements can reduce CPU time significantly.Furthermore the contribution of each improvement is defined here, and then their contributions are compared thoroughly.

Availability Analysis of Parallel Repairable System with Omitted Failures

The new model for parallel repairable system is introduced, and it is based on the practice problems of maintenance and the idea of Ion-Channel modeling. In the new model, repair times that are sufficiently short （less than some critical value） do not result in system failure, and such a repair interval is omitted from the downtime record. Usually, the underlying process is Markov process if the durations of working and repair time have the negative-exponential distributions, but the new system has not the Markov properties, which is worth to study. The reliability indexes such as instantaneous availability and steady-state availabilities for the new system are given through probability analysis. A numerical example is given to illustrate the results.

Stability of Difference Schemes with Intrinsic Parallelism for Quasilinear Parabolic Systems

In this paper, the first boundary problem of quasilinear parabolic system of second order is studied by the finite difference method with intrinsic parallelism. for the problem, the stability of the difference schemes with intrinsic parallelism are justified in the sense of the continuous dependence of the discrete vector solution of the difference schemes on the discrete data of the original problem, without assuming the existence of the smooth solutions for the origillal problem.

Reliability analysis of a two-dependent-unit parallel system with a standby unit

A parallel system with two active components and a cold standby unit is studied in this paper. The two simultaneously working components are dependent and the copula function is used to model their dependence. An explicit expression is obtained for the mean time to failure of the system in terms of the copula function and marginal lifetime distributions in two different cases. As an application,numerical calculations are presented corresponding to two different copula functions and marginal lifetime distributions.

A Class of Parallel Algorithms of Real-TimeNumerical Simulation for Stiff Dynamic System

In this paper a class of real-time parallel modified Rosenbrock methods of numerical simulation is constructed for stiff dynamic systems on a multiprocessor system, and convergence and numerical stability of these methods are discussed. A-stable real-time parallel formula of two-stage third-order and A(α)-stable real-time parallel formula with o ≈ 89.96° of three-stage fourth-order are particularly given. The numerical simulation experiments in parallel environment show that the class of algorithms is efficient and applicable, with greater speedup.

Improved Hungarian algorithm for assignment problems of serial-parallel systems

In order to overcome the shortcoming of the classical Hungarian algorithm that it can only solve the problems where the total cost is the sum of that of each job, an improved Hungarian algorithm is proposed and used to solve the assignment problem of serial-parallel systems. First of all, by replacing parallel jobs with virtual jobs, the proposed algorithm converts the serial-parallel system into a pure serial system, where the classical Hungarian algorithm can be used to generate a temporal assignment plan via optimization. Afterwards, the assignment plan is validated by checking whether the virtual jobs can be realized by real jobs through local searching. If the assignment plan is not valid, the converted system will be adapted by adjusting the parameters of virtual jobs, and then be optimized again. Through iterative searching, the valid optimal assignment plan can eventually be obtained.To evaluate the proposed algorithm, the valid optimal assignment plan is applied to labor allocation of a manufacturing system which is a typical serial-parallel system.

Reliability Comparison Between Parallel and Cold-standby Two-unit System Under Poisson Shock

This paper investigates the comparison problem of the reliability index between a parallel and a cold-standby system,both of which are consisting of two identical units.On the contrary to the general intuitive result,we proved that,under the condition that the system is shocked by a Poisson stream,the life time of the parallel system is longer than that of the cold-standby one in the sense of probability.

A Class of Parallel Runge-Kutta Methods for Differential-Algebraic Systems of Index 2

A class of parallel Runge-Kutta Methods for differential-algebraic equations of index 2are constructed for multiprocessor system. This paper gives the order conditions and investigatesthe convergence theory for such methods.

RELIABILITY ANALYSIS FOR A REPAIRABLE PARALLEL SYSTEM WITH TIME-VARYING FAILURE RATES

To solve a real problem:how to calculate the reliability of a system with time-varying failure rates in industry systems,this paper studies a model for the load-sharing parallel system with time-varying failure rates,and obtains calculating formulas of reliability and availability of the system by solving differential equations.In this paper,the failure rates are expressed in polynomial configuration.The constant,linear and Weibull failure rate are in their special form.The polynomial failure rates provide flexibility in modeling the practical time-varying failure rates.

Load Dependent Series-Parallel Systems with Common Bus Performance Sharing Mechanism

A series-parallel system was proposed with common bus performance sharing in which the performance and failure rate of the element depended on the load it was carrying. In such a system,the surplus performance of a sub-system can be transmitted to other deficient sub-systems. The transmission capacity of the common bus performance sharing mechanism is a random variable. Effects of load on element performance and failure rate were considered in this paper. A reliability evaluation algorithm based on the universal generating function technique was suggested. Numerical experiments were conducted to illustrate the algorithm.

Existence and Uniqueness of the Dynamic Solution of a Series-Parallel Repairable System Consisting of Three-Unit with Multiple Vacations of a Repairman

We investigate a series-parallel repairable system consisting of three-unit with multiple vacations of a repairman. By using C0-semigroup theory of linear operators in the functional analysis, we prove that the system is well-posed and has a unique positive dynamic solution.

MODELING AND PERFORMANCE ANALYSIS FOR THE SERIAL AND PARALLEL PRODUCTION SYSTEM BASED ON GSPN

Differed from the existed applications of generalized stochastic Petri net(GSPN) theory in machine-tool manufacturing system, reliability computation of FMS, testabilityparameters determination and fault analysis, a new idea of applying GSPN to model and performanceanalysis for the serial and parallel production system is proposed. And one typical discrete eventdynamic system (DEDS), turner-unit of palletizing system, is taken as a real case to research. Basedupon the established GSPN models, the working performances of serial and parallel layout arecompared. Furthermore, their differences of working mechanisms including feeding mechanism,coordinating mechanism and monitoring mechanism are discussed. Thus the theoretical basis which ishelpful to appraise layout plan and its reasonableness is provided. Meanwhile, the research resultsshow that parallel layout is more advantageous to greatly improve the operational speed ofproduction system than serial one.

A Parallel Multicomputer Simulation System

In this paper, a homogenous parallel simulation system is presented in detail for continuous--system simulation. The system is collstructed by a host computer and I I transputers connected into a topologyof 'Super--Node' which is very suitable for simulation of stiff systems. An automatic software interface runin the host is developed to partition simulation model, either equations or block diagrams, into several equitable segments and then pack them into parallel simulation program to be executed in the parallel system.This interface frees simulation users from parallel programming to focus on their simulation experiments.

Validation concurrency control protocol in parallel real-time database systems

In parallel real-time database systems, concurrency control protocols must satisfy time constraints as well as the integrity constraints. The authors present a validation concurrency control(VCC) protocol, which can enhance the performance of real-time concurrency control mechanism by reducing the number of transactions that might miss their deadlines, and compare the performance of validation concurrency control protocol with that of HP2PL(High priority two phase locking) protocol and OCC-TI-WAIT-50(Optimistic concurrency control-time interval-wait-50) protocol under shared-disk architecture by simulation. The simulation results reveal that the protocol the author presented can effectively reduce the number of transactions restarting which might miss their deadlines and performs better than HP2PL and OCC-TI-WAIT-50. It works well when arrival rate of transaction is lesser than threshold. However, due to resource contention the percentage of missing deadline increases sharply when arrival rate is greater than the threshold.

Multilayer Hex-Cells: A New Class of Hex-Cell Interconnection Networks for Massively Parallel Systems

Scalability is an important issue in the design of interconnection networks for massively parallel systems. In this paper a scalable class of interconnection network of Hex-Cell for massively parallel systems is introduced. It is called Multilayer Hex-Cell (MLH). A node addressing scheme and routing algorithm are also presented and discussed. An interesting feature of the proposed MLH is that it maintains a constant network degree regardless of the increase in the network size degree which facilitates modularity in building blocks of scalable systems. The new addressing node scheme makes the proposed routing algorithm simple and efficient in terms of that it needs a minimum number of calculations to reach the destination node. Moreover, the diameter of the proposed MLH is less than Hex-Cell network.

The Research on Stability of DC/DC in Parallel System

In this paper, a main structure of DC distributed power system is introduced;the stability of the output voltage in parallel system is put forward. This paper analyses the output impedance of master-slave current sharing mode and average current sharing, analyses the stability of parallel system through simulation, there is Right-half plane (RHP) polar in the Bode plot of input impedance. At last this paper distinguishes whether this system is stability and verifies the validity of the simulation.