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

Directional consistency of the overall cell population in the various scaled versions of the familiar 120 × 60 cm arena. Each plot depicts a differently stretched version as indicated by the corresponding diagrams in the top left corner of each plot.
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 22: Directional consistency of the overall cell population in the various scaled versions of the familiar 120 × 60 cm arena. Each plot depicts a differently stretched version as indicated by the corresponding diagrams in the top left corner of each plot.

Mentions: In O'Keefe and Burgess (1996) animals were trained to become familiar with a rectangular environment and then have their place cells being recorded from while exploring a number of differently scaled versions of the original arena. In their study, they reported a range of observations: fields were seen to stay at fixed distances from certain sections of the walls, stay at relative positions within the overall maze, and getting pulled apart when the maze was being stretched out. Figure 21 shows a number of results reported in O'Keefe and Burgess (1996) as well as a collection of selected cell activity as computed by our model, including one cell with a striking resemblance to the one reported in O'Keefe and Burgess (1996). The images shown were chosen to demonstrate the observed range of behavior, including the classes reported in O'Keefe and Burgess (1996). All 32 cells simulated in our model showed clear spatial activity in the original 120 × 60 cm setting as well as in the morphed versions—i.e., none of the cells broke down into an incoherent excess of activity, which usually happens if the model is unable to properly process its input. For almost all of the cells the activity in the morphed boxes can be labeled as either belonging to one of the categories reported in O'Keefe and Burgess (1996) or to one of the additional classes observed in our simulations: fields that rotated to stick with the same relative position to distinct landmarks (such as corners, for example); fields splitting into two, usually mirrored fields; and fields that relocated (often to the center of the new environment and usually observed in the square variants). While the firing patterns stayed coherent in almost all 32 cases, two units displayed an overall loss of localized firing, with different “mutations” of the original field in each of the stretch variants. Figures 22, 23 depict the statistics of this behavior for the overall population. While directional consistency is close to 1.0 for the familiar 120 × 60 cm environment, this value drops for most cells in the scaled versions of the arena, with the large 120 × 120 environment being the most confusing to the network. The number of firing fields behaves in a similar fashion, and cells can be seen to fall below 50% of their familiar-setting peak activity in the unfamiliar settings.


Modeling place field activity with hierarchical slow feature analysis.

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

Directional consistency of the overall cell population in the various scaled versions of the familiar 120 × 60 cm arena. Each plot depicts a differently stretched version as indicated by the corresponding diagrams in the top left corner of each plot.
© Copyright Policy - open-access
Related In: Results  -  Collection

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

Figure 22: Directional consistency of the overall cell population in the various scaled versions of the familiar 120 × 60 cm arena. Each plot depicts a differently stretched version as indicated by the corresponding diagrams in the top left corner of each plot.
Mentions: In O'Keefe and Burgess (1996) animals were trained to become familiar with a rectangular environment and then have their place cells being recorded from while exploring a number of differently scaled versions of the original arena. In their study, they reported a range of observations: fields were seen to stay at fixed distances from certain sections of the walls, stay at relative positions within the overall maze, and getting pulled apart when the maze was being stretched out. Figure 21 shows a number of results reported in O'Keefe and Burgess (1996) as well as a collection of selected cell activity as computed by our model, including one cell with a striking resemblance to the one reported in O'Keefe and Burgess (1996). The images shown were chosen to demonstrate the observed range of behavior, including the classes reported in O'Keefe and Burgess (1996). All 32 cells simulated in our model showed clear spatial activity in the original 120 × 60 cm setting as well as in the morphed versions—i.e., none of the cells broke down into an incoherent excess of activity, which usually happens if the model is unable to properly process its input. For almost all of the cells the activity in the morphed boxes can be labeled as either belonging to one of the categories reported in O'Keefe and Burgess (1996) or to one of the additional classes observed in our simulations: fields that rotated to stick with the same relative position to distinct landmarks (such as corners, for example); fields splitting into two, usually mirrored fields; and fields that relocated (often to the center of the new environment and usually observed in the square variants). While the firing patterns stayed coherent in almost all 32 cases, two units displayed an overall loss of localized firing, with different “mutations” of the original field in each of the stretch variants. Figures 22, 23 depict the statistics of this behavior for the overall population. While directional consistency is close to 1.0 for the familiar 120 × 60 cm environment, this value drops for most cells in the scaled versions of the arena, with the large 120 × 120 environment being the most confusing to the network. The number of firing fields behaves in a similar fashion, and cells can be seen to fall below 50% of their familiar-setting peak activity in the unfamiliar settings.

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