-
Towards Learning Stochastic Population Models by Gradient Descent
Authors:
Justin N. Kreikemeyer,
Philipp Andelfinger,
Adelinde M. Uhrmacher
Abstract:
Increasing effort is put into the development of methods for learning mechanistic models from data. This task entails not only the accurate estimation of parameters but also a suitable model structure. Recent work on the discovery of dynamical systems formulates this problem as a linear equation system. Here, we explore several simulation-based optimization approaches, which allow much greater fre…
▽ More
Increasing effort is put into the development of methods for learning mechanistic models from data. This task entails not only the accurate estimation of parameters but also a suitable model structure. Recent work on the discovery of dynamical systems formulates this problem as a linear equation system. Here, we explore several simulation-based optimization approaches, which allow much greater freedom in the objective formulation and weaker conditions on the available data. We show that even for relatively small stochastic population models, simultaneous estimation of parameters and structure poses major challenges for optimization procedures. Particularly, we investigate the application of the local stochastic gradient descent method, commonly used for training machine learning models. We demonstrate accurate estimation of models but find that enforcing the inference of parsimonious, interpretable models drastically increases the difficulty. We give an outlook on how this challenge can be overcome.
△ Less
Submitted 28 June, 2024; v1 submitted 10 April, 2024;
originally announced April 2024.
-
Automatic Reuse, Adaption, and Execution of Simulation Experiments via Provenance Patterns
Authors:
Pia Wilsdorf,
Anja Wolpers,
Jason Hilton,
Fiete Haack,
Adelinde M. Uhrmacher
Abstract:
Simulation experiments are typically conducted repeatedly during the model development process, for example, to re-validate if a behavioral property still holds after several model changes. Approaches for automatically reusing and generating simulation experiments can support modelers in conducting simulation studies in a more systematic and effective manner. They rely on explicit experiment speci…
▽ More
Simulation experiments are typically conducted repeatedly during the model development process, for example, to re-validate if a behavioral property still holds after several model changes. Approaches for automatically reusing and generating simulation experiments can support modelers in conducting simulation studies in a more systematic and effective manner. They rely on explicit experiment specifications and, so far, on user interaction for initiating the reuse. Thereby, they are constrained to support the reuse of simulation experiments in a specific setting. Our approach now goes one step further by automatically identifying and adapting the experiments to be reused for a variety of scenarios. To achieve this, we exploit provenance graphs of simulation studies, which provide valuable information about the previous modeling and experimenting activities, and contain meta-information about the different entities that were used or produced during the simulation study. We define provenance patterns and associate them with a semantics, which allows us to interpret the different activities, and construct transformation rules for provenance graphs. Our approach is implemented in a Reuse and Adapt framework for Simulation Experiments (RASE) which can interface with various modeling and simulation tools. In the case studies, we demonstrate the utility of our framework for a) the repeated sensitivity analysis of an agent-based model of migration routes, and b) the cross-validation of two models of a cell signaling pathway.
△ Less
Submitted 14 September, 2021;
originally announced September 2021.
-
An Artifact-based Workflow for Finite-Element Simulation Studies
Authors:
Andreas Ruscheinski,
Pia Wilsdorf,
Julius Zimmermann,
Ursula van Rienen,
Adelinde M. Uhrmacher
Abstract:
Workflow support typically focuses on single simulation experiments. This is also the case for simulation based on finite element methods. If entire simulation studies shall be supported, flexible means for intertwining revising the model, collecting data, executing and analyzing experiments are required. Artifact-based workflows present one means to support entire simulation studies, as has been…
▽ More
Workflow support typically focuses on single simulation experiments. This is also the case for simulation based on finite element methods. If entire simulation studies shall be supported, flexible means for intertwining revising the model, collecting data, executing and analyzing experiments are required. Artifact-based workflows present one means to support entire simulation studies, as has been shown for stochastic discrete-event simulation. To adapt the approach to finite element methods, the set of artifacts, i.e., conceptual model, requirement, simulation model, and simulation experiment, and the constraints that apply are extended by new artifacts, such as geometrical model, input data, and simulation data. Artifacts, their life cycles, and constraints are revisited revealing features both types of simulation studies share and those they vary in. Also, the potential benefits of exploiting an artifact-based workflow approach are shown based on a concrete simulation study. To those benefits belong guidance to systematically conduct simulation studies, reduction of effort by automatically executing specific steps, e.g., generating and executing convergence tests, and support for the automatic reporting of provenance.
△ Less
Submitted 15 October, 2020;
originally announced October 2020.
-
A flexible architecture for modeling and simulation of diffusional association
Authors:
Fiete Haack,
Stefan Leye,
Adelinde M. Uhrmacher
Abstract:
Up to now, it is not possible to obtain analytical solutions for complex molecular association processes (e.g. Molecule recognition in Signaling or catalysis). Instead Brownian Dynamics (BD) simulations are commonly used to estimate the rate of diffusional association, e.g. to be later used in mesoscopic simulations. Meanwhile a portfolio of diffusional association (DA) methods have been develop…
▽ More
Up to now, it is not possible to obtain analytical solutions for complex molecular association processes (e.g. Molecule recognition in Signaling or catalysis). Instead Brownian Dynamics (BD) simulations are commonly used to estimate the rate of diffusional association, e.g. to be later used in mesoscopic simulations. Meanwhile a portfolio of diffusional association (DA) methods have been developed that exploit BD.
However, DA methods do not clearly distinguish between modeling, simulation, and experiment settings. This hampers to classify and compare the existing methods with respect to, for instance model assumptions, simulation approximations or specific optimization strategies for steering the computation of trajectories.
To address this deficiency we propose FADA (Flexible Architecture for Diffusional Association) - an architecture that allows the flexible definition of the experiment comprising a formal description of the model in SpacePi, different simulators, as well as validation and analysis methods. Based on the NAM (Northrup-Allison-McCammon) method, which forms the basis of many existing DA methods, we illustrate the structure and functioning of FADA. A discussion of future validation experiments illuminates how the FADA can be exploited in order to estimate reaction rates and how validation techniques may be applied to validate additional features of the model.
△ Less
Submitted 22 February, 2010;
originally announced February 2010.
