Design Principles of the Hippocampal Cognitive Map

By Kimberly L. Stachenfeld, Matthew Botvinick, Samuel J. Gershman. 2014

114 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation (2016, background): “Decomposition of the successor representation have shown to yield reasonable subgoals for spatial navigation problems [5, 30].”
Deep Successor Reinforcement Learning (2016, background): “2”
Toward an Integration of Deep Learning and Neuroscience (2016, background): “…and it has been suggested that error-driven learning of an optimal such representation in the hippocampus 50 gives rise to place and grid cell properties (Stachenfeld, 2014), with goal representations themselves emerging in the amygdala, prefrontal cortex and other areas (O’Reilly et al., 2014a).”
A map of abstract relational knowledge in the human hippocampal–entorhinal cortex (2017, background): “However, recent theoretical treatments suggest that grid-like firing might also be understood as a principal component decomposition of the covariance information between place cells (Dordek et al., 2016) or the transitions between states (Stachenfeld et al., 2016, 2014).”
Reinforcement Learning and Causal Models (2017, background): “In particular, Stachenfeld, Botvinick, and Gershman (2014) showed that one way to construct a cognitive map is to learn a predictive representation (Dayan, 1993), which is, in essence, an association between current and future states.”
Eigenoption Discovery through the Deep Successor Representation (2017, methodology): “To the best of our knowledge, this equivalence was ﬁrst explicitly discussed by Stachenfeld et al. (2014).”
Neuroscience-Inspired Artificial Intelligence. (2017, background): “Grid codes may be an excellent candidate for organizing conceptual knowledge, because they allow state spaces to be decomposed efﬁciently, in a way that could support discovery of subgoals and hierarchical planning (Stachenfeld et al., 2014).”
Organizing Conceptual Knowledge in Humans with a Grid-like Code (2016, background): “problems that are not easily mapped onto continuous 2-dimensional spaces(27, 28).”
If deep learning is the answer, what is the question? (2020, background): “Neuroscientists have even begun to unravel how seemingly idiosyncratic coding properties in the MTL and other structures may be hallmarks of a normative scheme for computationally efficient planning and inference 110,111 .”
A Geometric Perspective on Optimal Representations for Reinforcement Learning (2019, methodology): “largest-eigenvalue eigenvectors of the Laplacian of the transition function’s graphical structure, recent formulations use the top singular vectors of (I Pˇ) 1, where ˇis the uniformly random policy (Stachenfeld et al., 2014; Machado et al., 2017; Behzadian and Petrik, 2018). In line with the analys”
The successor representation in human reinforcement learning (2016, background): “…suggests that the SR is explicitly represented in the brain: the SR defined over space can capture many properties of rodent hippocampal place cells 11 , whereas in tasks with more abstract sequential stimulus structure, the SR captures properties of fMRI pattern similarity in hippocampal and…”
The Laplacian in RL: Learning Representations with Efficient Approximations (2018, methodology): “d to approximate the Laplacian eigenvectors with functions approximators via an equivalence between proto-value functions (Mahadevan, 2005) and spectral decomposition of the successor representation (Stachenfeld et al., 2014). Importantly, they propose an approach for stochastically approximating th”
Hippocampal Contributions to Model-Based Planning and Spatial Memory (2019, unknown): “The related successor representation model (Stachenfeld et al., 2014; Garvert et al., 2017) also focuses on learned relationships among locations by proposing that place selectivity itself is built from experience of state transitions to reflect expectations about future locations.”
Intracranial Recordings Demonstrate Both Cortical and Medial Temporal Lobe Engagement in Visual Search in Humans (2021, methodology): “We did not see increased MTL activity during the sample interval, as would have been expected if the MTL prospectively guides top– down attention signals by increasing activity during cue presentation (Stokes et al., 2012).”