Elements of a stochastic 3D prediction engine in larval zebrafish prey capture
- Harvard University, United States
- Massachusetts Institute of Technology, United States
Abstract
The computational principles underlying predictive capabilities in animals are poorly understood. Here, we wondered whether predictive models mediating prey capture could be reduced to a simple set of sensorimotor rules performed by a primitive organism. For this task, we chose the larval zebrafish, a tractable vertebrate that pursues and captures swimming microbes. Using a novel naturalistic 3D setup, we show that the zebrafish combines position and velocity perception to construct a future positional estimate of its prey, indicating an ability to project trajectories forward in time. Importantly, the stochasticity in the fish's sensorimotor transformations provides a considerable advantage over equivalent noise-free strategies. This surprising result coalesces with recent findings that illustrate the benefits of biological stochasticity to adaptive behavior. In sum, our study reveals that zebrafish are equipped with a recursive prey capture algorithm, built up from simple stochastic rules, that embodies an implicit predictive model of the world.
Data availability
All software related to behavioral analysis, modeling, and virtual prey capture simulation is freely available at www.github.com/larrylegend33/PreycapMaster. The software is licensed under a GNU General Public License 3.0. Source data for analysis and simulations is enclosed as "Source Data" in relevant figures. Source Data for Figure 2 contains all pursuit bouts analyzed in the dataset; it was used to construct Figures 2, 3, 5, and 6A, and is accompanied by instructions for running queries. Source Data for Figure 6 contains the generators for simulating from the DPMMs in Figure 6. Using the code at www.github.com/larrylegend33/PreycapMaster and the generators in Source Data - Figure 6 requires obtaining the BayesDB software package, which is freely available at http://probcomp.csail.mit.edu/software/bayesdb/.
Article and author information
Author details
Funding
National Institutes of Health (U19NS104653)
- Florian Engert
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Ethics
Animal experimentation: Experiments were conducted according to the guidelines of the National Institutes of Health and were approved by the Standing Committee on the Use of Animals in Research of Harvard University. Animals were handled according IACUC protocol #2729.
Copyright
© 2019, Bolton 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
-
- 3,466
- views
-
- 410
- downloads
-
- 33
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
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)
Elements of a stochastic 3D prediction engine in larval zebrafish prey capture
eLife 8:e51975.
https://doi.org/10.7554/eLife.51975
Further reading
-
- Neuroscience
The relation between neural activity and behaviorally relevant variables is at the heart of neuroscience research. When strong, this relation is termed a neural representation. There is increasing evidence, however, for partial dissociations between activity in an area and relevant external variables. While many explanations have been proposed, a theoretical framework for the relationship between external and internal variables is lacking. Here, we utilize recurrent neural networks (RNNs) to explore the question of when and how neural dynamics and the network’s output are related from a geometrical point of view. We find that training RNNs can lead to two dynamical regimes: dynamics can either be aligned with the directions that generate output variables, or oblique to them. We show that the choice of readout weight magnitude before training can serve as a control knob between the regimes, similar to recent findings in feedforward networks. These regimes are functionally distinct. Oblique networks are more heterogeneous and suppress noise in their output directions. They are furthermore more robust to perturbations along the output directions. Crucially, the oblique regime is specific to recurrent (but not feedforward) networks, arising from dynamical stability considerations. Finally, we show that tendencies toward the aligned or the oblique regime can be dissociated in neural recordings. Altogether, our results open a new perspective for interpreting neural activity by relating network dynamics and their output.
-
- Neuroscience
In this paper, we provide an overview and analysis of the BRAIN Initiative data-sharing ecosystem. First, we compare and contrast the characteristics of the seven BRAIN Initiative data archives germane to data sharing and reuse, namely data submission and access procedures and aspects of interoperability. Second, we discuss challenges, benefits, and future opportunities, focusing on issues largely specific to sharing human data and drawing on N = 34 interviews with diverse stakeholders. The BRAIN Initiative-funded archive ecosystem faces interoperability and data stewardship challenges, such as achieving and maintaining interoperability of data and archives and harmonizing research participants’ informed consents for tiers of access for human data across multiple archives. Yet, a benefit of this distributed archive ecosystem is the ability of more specialized archives to adapt to the needs of particular research communities. Finally, the multiple archives offer ample raw material for network evolution in response to the needs of neuroscientists over time. Our first objective in this paper is to provide a guide to the BRAIN Initiative data-sharing ecosystem for readers interested in sharing and reusing neuroscience data. Second, our analysis supports the development of empirically informed policy and practice aimed at making neuroscience data more findable, accessible, interoperable, and reusable.
