Computational optimization of associative learning experiments

Abstract

With computational biology striving to provide more accurate theoretical accounts of biological systems, use of increasingly complex computational models seems inevitable. However, this trend engenders a challenge of optimal experimental design: due to the flexibility of complex models, it is difficult to intuitively design experiments that will efficiently expose differences between candidate models or allow accurate estimation of their parameters. This challenge is well exemplified in associative learning research. Associative learning theory has a rich tradition of computational modeling, resulting in a growing space of increasingly complex models, which in turn renders manual design of informative experiments difficult. Here we propose a novel method for computational optimization of associative learning experiments. We first formalize associative learning experiments using a low number of tunable design variables, to make optimization tractable. Next, we combine simulation-based Bayesian experimental design with Bayesian optimization to arrive at a flexible method of tuning design variables. Finally, we validate the proposed method through extensive simulations covering both the objectives of accurate parameter estimation and model selection. The validation results show that computationally optimized experimental designs have the potential to substantially improve upon manual designs drawn from the literature, even when prior information guiding the optimization is scarce. Computational optimization of experiments may help address recent concerns over reproducibility by increasing the expected utility of studies, and it may even incentivize practices such as study pre-registration, since optimization requires a pre-specified analysis plan. Moreover, design optimization has the potential not only to improve basic research in domains such as associative learning, but also to play an important role in translational research. For example, design of behavioral and physiological diagnostic tests in the nascent field of computational psychiatry could benefit from an optimization-based approach, similar to the one presented here.

Measuring learning in human classical threat conditioning: Translational, cognitive and methodological considerations

“…To what these quantities differ between CS+ and CS− trials can depend on the experimental design, such that some designs may favor one CR and other designs a different CR. Regarding the influence of experimental design on inference in associative learning experiments, we refer the reader to Melinscak and Bach (2020).…”

Inconsistent analytic strategies reduce robustness in fear extinction via skin conductance response

“…These approaches rely on increased transparency in data reporting and analysis, and we maintain that decisions during data reduction and analysis should be reported and justified (Lonsdorf, Klingelhöfer‐Jens, et al, 2019; Ney et al., 2018). It is also possible that reproducibility may be improved by computational optimization of paradigm design, which may aid in determining how to vary experimental parameters to answer specific research questions (Melinscak & Bach, 2020).…”

“…Similarly, the cause for inadequate robustness between trial‐by‐trial and averaged data should be systematically investigated. It is possible that the failure of these methods to replicate is due to lack of power, in which case methods that seek to improve power via experimental design and SCR scoring are highly desirable (Bach & Melinscak, 2020; Melinscak & Bach, 2020). Specifically, improving SCR scoring can provide better estimates of responses relevant to extinction paradigms by reducing measurement error (Bach & Melinscak, 2020), and optimization of experimental designs based on statistical requirements can make analyses more amenable to experimental data (Melinscak & Bach, 2020).…”

Improving the Reliability of Cognitive Task Measures: A Narrative Review

“…Estimation noise is the inverse of the (relative or absolute) agreement between the true and recovered parameters. Alterations to experimental design can improve parameter recovery and estimation noise, and multiple frameworks have been proposed for testing and improving experimental designs to aid parameter recovery [63,64]. Parameter recovery can also be affected by the model estimation method [34,65].…”