More Like this

1. Bounded Continuous-Time Satisfiability Solver

   Yamashita, H. ; Suzuki, H. ; Toroczkai, Z. ; Aihara, K. ( December 2019 , International Symposium on Nonlinear Theory and its Applications (NOLTA2019))

   To tackle problems that can not be solved by current digital computers, many systems propose ideas from physics and neuroscience. The CTDS solver introduced by Ercsey-Ravasz and Toroczkai is one of such system. It solves the satisfiability problem by reducing it to a minimization of a time-varying target function. Although the possibility of an efficient electric circuit implementation of the solver has been shown, in terms of physical realizations, the solver has a problem of unbounded variations of the target function parameters. Here we propose a variant of the solver with bounded target function parameters. It includes several possible modifications of the solver in system parameter differences. We also show the basic characteristics of the solver, the upper and lower bounds of the target function parameters. 

   more » « less
2. OVERCOMING CATASTROPHIC FORGETTING FOR CONTINUAL LEARNING VIA MODEL ADAPTATION

   Hu, Wenpeng ; Lin, Zhou ; Liu, Bing ; Tao, Chongyang ; Tao, Zhengwei Tao ; Zhao, Dongyan ; Ma, Jinwen ; Yan, Rui ( May 2019 , International Conference on Learning Representations)

   Learning multiple tasks sequentially is important for the development of AI and lifelong learning systems. However, standard neural network architectures suffer from catastrophic forgetting which makes it difficult for them to learn a sequence of tasks. Several continual learning methods have been proposed to address the problem. In this paper, we propose a very different approach, called Parameter Generation and Model Adaptation (PGMA), to dealing with the problem. The proposed approach learns to build a model, called the solver, with two sets of parameters. The first set is shared by all tasks learned so far and the second set is dynamically generated to adapt the solver to suit each test example in order to classify it. Extensive experiments have been carried out to demonstrate the effectiveness of the proposed approach. 

   more » « less
3. Topology Optimization Design of Structures Based on Eigenfrequency Matching

   https://doi.org/10.1115/DETC2021-69498  

   Giraldo-Guzmán, Daniel ; Lissenden, Clifford ; Shokouhi, Parisa ; Frecker, Mary ( January 2021 , 47th Design Automation Conference (DAC))

   We demonstrate the design of resonating structures using a density-based topology optimization approach, which requires the eigenfrequencies to match a set of target values. To develop a solution, several optimization modules are implemented, including material interpolation models, penalization schemes, filters, analytical sensitivities, and a solver. Moreover, common challenges in topology optimization for dynamic systems and their solutions are discussed. In this study, the objective function is to minimize the error between the target and actual eigenfrequency values. The finite element method is used to compute the eigenfrequencies at each iteration. To solve the optimization problem, we use the sequential linear programming algorithm with move limits, enhanced by a filtering technique. Finally, we present a resonator design as a case study and analyze the design process with different optimization parameters.

    

   more » « less
4. Data Farming the Parameters of Simulation-Optimization Solvers

   https://doi.org/10.1145/3680282  

   Shashaani, Sara ; Eckman, David ; Sanchez, Susan ( October 2024 , ACM Transactions on Modeling and Computer Simulation)

   The performance of a simulation-optimization algorithm, a.k.a. a solver, depends on its parameter settings. Much of the research to date has focused on how a solver’s parameters affect its convergence and other asymptotic behavior. While these results are important for providing a theoretical understanding of a solver, they can be of limited utility to a user who must set up and run the solver on a particular problem. When running a solver in practice, good finite-time performance is paramount. In this article, we explore the relationship between a solver’s parameter settings and its finite-time performance by adopting a data farming approach. The approach involves conducting and analyzing the outputs of a designed experiment wherein the factors are the solver’s parameters and the responses are assorted performance metrics measuring the solver’s speed and solution quality over time. We demonstrate this approach with a study of the ASTRO-DF solver when solving a stochastic activity network problem and an inventory control problem. Through these examples, we show that how some of the solver’s parameters are set greatly affects its ability to achieve rapid, reliable progress and gain insights into the solver’s inner workings. We discuss the implications of using this framework for tuning solver parameters, as well as for addressing related questions of interest to solver specialists and generalists. 

   more » « less
5. Decomposed Iterative Optimal Power Flow with Automatic Regionalization

   https://doi.org/10.3390/en13184987  

   Zheng, Xinhu ; Duan, Dongliang ; Yang, Liuqing ; Wang, Haonan ( September 2020 , Energies)

   The optimal power flow (OPF) problem plays an important role in power system operation and control. The problem is nonconvex and NP-hard, hence global optimality is not guaranteed and the complexity grows exponentially with the size of the system. Therefore, centralized optimization techniques are not suitable for large-scale systems and an efficient decomposed implementation of OPF is highly demanded. In this paper, we propose a novel and efficient method to decompose the entire system into multiple sub-systems based on automatic regionalization and acquire the OPF solution across sub-systems via a modified MATPOWER solver. The proposed method is implemented in a modified solver and tested on several IEEE Power System Test Cases. The performance is shown to be more appealing compared with the original solver. 

   more » « less