Refine
Document Type
- Working Paper (2) (remove)
Language
- English (2)
Has Fulltext
- yes (2) (remove)
Keywords
- Simulation (2) (remove)
Benchmark experiments are required to test, compare, tune, and understand optimization algorithms. Ideally, benchmark problems closely reflect real-world problem behavior. Yet, real-world problems are not always readily available for benchmarking. For example, evaluation costs may be too high, or resources are unavailable (e.g., software or equipment). As a solution, data from previous evaluations can be used to train surrogate models which are then used for benchmarking. The goal is to generate test functions on which the performance of an algorithm is similar to that on the real-world objective function. However, predictions from data-driven models tend to be smoother than the ground-truth from which the training data is derived. This is especially problematic when the training data becomes sparse. The resulting benchmarks may not reflect the landscape features of the ground-truth, are too easy, and may lead to biased conclusions.
To resolve this, we use simulation of Gaussian processes instead of estimation (or prediction). This retains the covariance properties estimated during model training. While previous research suggested a decomposition-based approach for a small-scale, discrete problem, we show that the spectral simulation method enables simulation for continuous optimization problems. In a set of experiments with an artificial ground-truth, we demonstrate that this yields more accurate benchmarks than simply predicting with the Gaussian process model.
When designing or developing optimization algorithms, test functions are crucial to evaluate
performance. Often, test functions are not sufficiently difficult, diverse, flexible or relevant to real-world
applications. Previously,
test functions with real-world relevance were generated by training a machine learning model based on
real-world data. The model estimation is used as a test function.
We propose a more principled approach using simulation instead of estimation.
Thus, relevant and varied test functions
are created which represent the behavior of real-world fitness landscapes.
Importantly, estimation can lead to excessively smooth test functions
while simulation may avoid this pitfall. Moreover, the simulation
can be conditioned by the data, so that the simulation reproduces the training data
but features diverse behavior in unobserved regions of the search space.
The proposed test function generator is illustrated with an intuitive, one-dimensional
example. To demonstrate the utility of this approach it
is applied to a protein sequence optimization problem.
This application demonstrates the advantages as well as practical limits of simulation-based
test functions.