Limits...
Modeling place field activity with hierarchical slow feature analysis.

Schönfeld F, Wiskott L - Front Comput Neurosci (2015)

Bottom Line: Simulations are based on a hierarchical Slow Feature Analysis (SFA) network topped by a principal component analysis (ICA) output layer.The slowness principle is shown to account for the main findings of the presented experimental studies.Future iterations of the model will thus have to incorporate additional information, such as path integration and grid cell activity, in order to be able to also replicate studies that take place during darkness.

View Article: PubMed Central - PubMed

Affiliation: Theory of Neural Systems Group, Institut für Neuroinformatik, Ruhr Universität Bochum Bochum, Germany.

ABSTRACT
What are the computational laws of hippocampal activity? In this paper we argue for the slowness principle as a fundamental processing paradigm behind hippocampal place cell firing. We present six different studies from the experimental literature, performed with real-life rats, that we replicated in computer simulations. Each of the chosen studies allows rodents to develop stable place fields and then examines a distinct property of the established spatial encoding: adaptation to cue relocation and removal; directional dependent firing in the linear track and open field; and morphing and scaling the environment itself. Simulations are based on a hierarchical Slow Feature Analysis (SFA) network topped by a principal component analysis (ICA) output layer. The slowness principle is shown to account for the main findings of the presented experimental studies. The SFA network generates its responses using raw visual input only, which adds to its biological plausibility but requires experiments performed in light conditions. Future iterations of the model will thus have to incorporate additional information, such as path integration and grid cell activity, in order to be able to also replicate studies that take place during darkness.

No MeSH data available.


Related in: MedlinePlus

Screenshots of our software tool. (A) Rendered overview of a simulated square environment featuring a white cue card and the position of the virtual animal marked by a black arrow. (B) Exemplary image frame produced by the software; a sequence of such 320 by 40 pixel frames is used to train our network. (C) Trajectory of a virtual rat running for 10 simulated minutes within a circular environment. (D) Trajectory of a real rat exploring a circular environment of the same radius for 10 min of real time. Note that the virtual agent runs at a constant speed of 20 cm/s and does not slow down or stop as real rats do. Thus, simulated rats tend to cover more ground than their real-life counterparts in the same amount of time.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4441153&req=5

Figure 2: Screenshots of our software tool. (A) Rendered overview of a simulated square environment featuring a white cue card and the position of the virtual animal marked by a black arrow. (B) Exemplary image frame produced by the software; a sequence of such 320 by 40 pixel frames is used to train our network. (C) Trajectory of a virtual rat running for 10 simulated minutes within a circular environment. (D) Trajectory of a real rat exploring a circular environment of the same radius for 10 min of real time. Note that the virtual agent runs at a constant speed of 20 cm/s and does not slow down or stop as real rats do. Thus, simulated rats tend to cover more ground than their real-life counterparts in the same amount of time.

Mentions: The overall software framework offers a pipeline of four distinct steps that can be interchanged and concatenated to replicate a wide range of experimental studies (cf. Schönfeld and Wiskott, 2013). Once the setup is defined by the user, the software places the virtual rat within the maze and records the visual stream of the agent while it performs the predefined experiment. The recorded data is then used to train a hierarchical SFA network as described above. The trained network is then sampled over the whole environment and automatically generates the plots presented throughout this work. To replicate the chosen experimental studies we reconstructed the respective environments as closely as the software permits. Maze layouts are matched to scale, similar textures and cue cards are being used, and the background consists of either a closed curtain or an office panorama at a distance (resulting in a realistic parallax). Since the simulated field of view is very narrow in the vertical dimension, however, wall height is usually set to a lower value in order to allow the agent to peek over the walls just like its real life counterpart. Figure 2 shows an example screenshot from our software after concluding a simulation run; it includes a graph of the trajectory of the artificial animal as well an exemplary image frame depicting the view of the agent.


Modeling place field activity with hierarchical slow feature analysis.

Schönfeld F, Wiskott L - Front Comput Neurosci (2015)

Screenshots of our software tool. (A) Rendered overview of a simulated square environment featuring a white cue card and the position of the virtual animal marked by a black arrow. (B) Exemplary image frame produced by the software; a sequence of such 320 by 40 pixel frames is used to train our network. (C) Trajectory of a virtual rat running for 10 simulated minutes within a circular environment. (D) Trajectory of a real rat exploring a circular environment of the same radius for 10 min of real time. Note that the virtual agent runs at a constant speed of 20 cm/s and does not slow down or stop as real rats do. Thus, simulated rats tend to cover more ground than their real-life counterparts in the same amount of time.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4441153&req=5

Figure 2: Screenshots of our software tool. (A) Rendered overview of a simulated square environment featuring a white cue card and the position of the virtual animal marked by a black arrow. (B) Exemplary image frame produced by the software; a sequence of such 320 by 40 pixel frames is used to train our network. (C) Trajectory of a virtual rat running for 10 simulated minutes within a circular environment. (D) Trajectory of a real rat exploring a circular environment of the same radius for 10 min of real time. Note that the virtual agent runs at a constant speed of 20 cm/s and does not slow down or stop as real rats do. Thus, simulated rats tend to cover more ground than their real-life counterparts in the same amount of time.
Mentions: The overall software framework offers a pipeline of four distinct steps that can be interchanged and concatenated to replicate a wide range of experimental studies (cf. Schönfeld and Wiskott, 2013). Once the setup is defined by the user, the software places the virtual rat within the maze and records the visual stream of the agent while it performs the predefined experiment. The recorded data is then used to train a hierarchical SFA network as described above. The trained network is then sampled over the whole environment and automatically generates the plots presented throughout this work. To replicate the chosen experimental studies we reconstructed the respective environments as closely as the software permits. Maze layouts are matched to scale, similar textures and cue cards are being used, and the background consists of either a closed curtain or an office panorama at a distance (resulting in a realistic parallax). Since the simulated field of view is very narrow in the vertical dimension, however, wall height is usually set to a lower value in order to allow the agent to peek over the walls just like its real life counterpart. Figure 2 shows an example screenshot from our software after concluding a simulation run; it includes a graph of the trajectory of the artificial animal as well an exemplary image frame depicting the view of the agent.

Bottom Line: Simulations are based on a hierarchical Slow Feature Analysis (SFA) network topped by a principal component analysis (ICA) output layer.The slowness principle is shown to account for the main findings of the presented experimental studies.Future iterations of the model will thus have to incorporate additional information, such as path integration and grid cell activity, in order to be able to also replicate studies that take place during darkness.

View Article: PubMed Central - PubMed

Affiliation: Theory of Neural Systems Group, Institut für Neuroinformatik, Ruhr Universität Bochum Bochum, Germany.

ABSTRACT
What are the computational laws of hippocampal activity? In this paper we argue for the slowness principle as a fundamental processing paradigm behind hippocampal place cell firing. We present six different studies from the experimental literature, performed with real-life rats, that we replicated in computer simulations. Each of the chosen studies allows rodents to develop stable place fields and then examines a distinct property of the established spatial encoding: adaptation to cue relocation and removal; directional dependent firing in the linear track and open field; and morphing and scaling the environment itself. Simulations are based on a hierarchical Slow Feature Analysis (SFA) network topped by a principal component analysis (ICA) output layer. The slowness principle is shown to account for the main findings of the presented experimental studies. The SFA network generates its responses using raw visual input only, which adds to its biological plausibility but requires experiments performed in light conditions. Future iterations of the model will thus have to incorporate additional information, such as path integration and grid cell activity, in order to be able to also replicate studies that take place during darkness.

No MeSH data available.


Related in: MedlinePlus