Difference between revisions of "Project 20 Benchmark Problems for Modelling Intracellular Processes"

(20 Benchmark Problem for Modelling Intracellular Processes)
Line 1: Line 1:
== 20 Benchmark Problem for Modelling Intracellular Processes ==
+
== Summary ==
 
 
 
Ordinary differential equation (ODE) models are frequently applied to describe intracellular biochemical processes.  
 
Ordinary differential equation (ODE) models are frequently applied to describe intracellular biochemical processes.  
 
Here, we provide a set of 20 published benchmark models which should serve as test cases for assessing methodology for modelling.
 
Here, we provide a set of 20 published benchmark models which should serve as test cases for assessing methodology for modelling.
  
 +
An overview about the benchmark models is also provided at the github repository [REF].
  
=== Download ===
+
== Download ==
 
 
 
The 20 benchmark models are available with version control on github [REF].
 
The 20 benchmark models are available with version control on github [REF].
  
=== Overview ===
 
 
An overview about the benchmark models is also provided at the github repository [REF].
 
 
=== Bechmarking Studies ===
 
  
 +
== Studies based on the 20 Benchmark problems ==
 
Here, we summarize several outcomes of benchmark studies performed on the 20 benchmark problems.
 
Here, we summarize several outcomes of benchmark studies performed on the 20 benchmark problems.
  
==== Optimization at the Log-Scale ====
+
=== Optimization at the Log-Scale ===
 
It has been shown [REF] that optimization algorithms have superior performance if parameters are optimized at the log-scale [REF].
 
It has been shown [REF] that optimization algorithms have superior performance if parameters are optimized at the log-scale [REF].
  
==== Convexity ====
+
=== Convexity ===
 
It has been shown [REF] that the negative log-likelihood which has to be minimized for parameter estimation is more convex if the parameters are evaluated at the log-scale. It has been concluded that this fact is one reason for performance benefits of deterministic optimization algorithms if optimization is performed at the log-scale [REF].
 
It has been shown [REF] that the negative log-likelihood which has to be minimized for parameter estimation is more convex if the parameters are evaluated at the log-scale. It has been concluded that this fact is one reason for performance benefits of deterministic optimization algorithms if optimization is performed at the log-scale [REF].

Revision as of 14:39, 9 August 2018

Summary

Ordinary differential equation (ODE) models are frequently applied to describe intracellular biochemical processes. Here, we provide a set of 20 published benchmark models which should serve as test cases for assessing methodology for modelling.

An overview about the benchmark models is also provided at the github repository [REF].

Download

The 20 benchmark models are available with version control on github [REF].


Studies based on the 20 Benchmark problems

Here, we summarize several outcomes of benchmark studies performed on the 20 benchmark problems.

Optimization at the Log-Scale

It has been shown [REF] that optimization algorithms have superior performance if parameters are optimized at the log-scale [REF].

Convexity

It has been shown [REF] that the negative log-likelihood which has to be minimized for parameter estimation is more convex if the parameters are evaluated at the log-scale. It has been concluded that this fact is one reason for performance benefits of deterministic optimization algorithms if optimization is performed at the log-scale [REF].