Flexible and efficient simulation-based inference for models of decision-making
Abstract
Inferring parameters of computational models that capture experimental data is a central task in cognitive neuroscience. Bayesian statistical inference methods usually require the ability to evaluate the likelihood of the model—however, for many models of interest in cognitive neuroscience, the associated likelihoods cannot be computed efficiently. Simulation-based inference (SBI) offers a solution to this problem by only requiring access to simulations produced by the model. Previously, Fengler et al. introduced Likelihood Approximation Networks (LAN, Fengler et al., 2021) which make it possible to apply SBI to models of decision-making, but require billions of simulations for training. Here, we provide a new SBI method that is substantially more simulation-efficient. Our approach, Mixed Neural Likelihood Estimation (MNLE), trains neural density estimators on model simulations to emulate the simulator, and is designed to capture both the continuous (e.g., reaction times) and discrete (choices) data of decision-making models. The likelihoods of the emulator can then be used to perform Bayesian parameter inference on experimental data using standard approximate inference methods like Markov Chain Monte Carlo sampling. We demonstrate MNLE on two variants of the drift-diffusion model (DDM) and show that it is substantially more efficient than LANs: MNLE achieves similar likelihood accuracy with six orders of magnitude fewer training simulations, and is significantly more accurate than LANs when both are trained with the same budget. This enables researchers to perform SBI on custom-tailored models of decision-making, leading to fast iteration of model design for scientific discovery.
Data availability
We implemented MNLE as part of the open source package for SBI, sbi, available at https://github. com/mackelab/sbi. Code for reproducing the results presented here, and tutorials on how to apply MNLE to other simulators using sbi can be found at https://github.com/mackelab/mnle-for-ddms.
Article and author information
Author details
Funding
Deutsche Forschungsgemeinschaft (SFB 1233)
- Jan-Matthis Lueckmann
- Jakob H Macke
Deutsche Forschungsgemeinschaft (SPP 2041)
- Jan Boelts
- Jakob H Macke
Deutsche Forschungsgemeinschaft (Germany's Excellence Strategy MLCoE)
- Jan Boelts
- Jan-Matthis Lueckmann
- Richard Gao
- Jakob H Macke
Bundesministerium für Bildung und Forschung (ADIMEM,FKZ 01IS18052 A-D)
- Jan-Matthis Lueckmann
- Jakob H Macke
HORIZON EUROPE Marie Sklodowska-Curie Actions (101030918)
- Richard Gao
Bundesministerium für Bildung und Forschung (Tübingen AI Center,FKZ 01IS18039A)
- Jan Boelts
- Jakob H Macke
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Copyright
© 2022, Boelts et al.
This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 3,613
- views
-
- 703
- downloads
-
- 26
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Neuroscience
Astrocytes derive from different lineages and play a critical role in neuropathic pain after spinal cord injury (SCI). Whether selectively eliminating these main origins of astrocytes in lumbar enlargement could attenuate SCI-induced neuropathic pain remains unclear. Through transgenic mice injected with an adeno-associated virus vector and diphtheria toxin, astrocytes in lumbar enlargement were lineage traced, targeted, and selectively eliminated. Pain-related behaviors were measured with an electronic von Frey apparatus and a cold/hot plate after SCI. RNA sequencing, bioinformatics analysis, molecular experiment, and immunohistochemistry were used to explore the potential mechanisms after astrocyte elimination. Lineage tracing revealed that the resident astrocytes but not ependymal cells were the main origins of astrocytes-induced neuropathic pain. SCI-induced mice to obtain significant pain symptoms and astrocyte activation in lumbar enlargement. Selective resident astrocyte elimination in lumbar enlargement could attenuate neuropathic pain and activate microglia. Interestingly, the type I interferons (IFNs) signal was significantly activated after astrocytes elimination, and the most activated Gene Ontology terms and pathways were associated with the type I IFNs signal which was mainly activated in microglia and further verified in vitro and in vivo. Furthermore, different concentrations of interferon and Stimulator of interferon genes (STING) agonist could activate the type I IFNs signal in microglia. These results elucidate that selectively eliminating resident astrocytes attenuated neuropathic pain associated with type I IFNs signal activation in microglia. Targeting type I IFNs signals is proven to be an effective strategy for neuropathic pain treatment after SCI.
-
- Neuroscience
Systems consolidation theories propose two mechanisms that enable the behavioral integration of related memories: coordinated reactivation between hippocampus and cortex, and the emergence of cortical traces that reflect overlap across memories. However, there is limited empirical evidence that links these mechanisms to the emergence of behavioral integration over time. In two experiments, participants implicitly encoded sequences of objects with overlapping structure. Assessment of behavioral integration showed that response times during a recognition task reflected behavioral priming between objects that never occurred together in time but belonged to overlapping sequences. This priming was consolidation-dependent and only emerged for sequences learned 24 hr prior to the test. Critically, behavioral integration was related to changes in neural pattern similarity in the medial prefrontal cortex and increases in post-learning rest connectivity between the posterior hippocampus and lateral occipital cortex. These findings suggest that memories with a shared predictive structure become behaviorally integrated through a consolidation-related restructuring of the learned sequences, providing insight into the relationship between different consolidation mechanisms that support behavioral integration.