RUB » Mercator Research Groups » Structure of memory »

Potential Bachelor/ Master Theses

Modeling enhanced replay of neuronal sequences in the hippocampus

A large body of experimental evidence shows that place cells in the hippocampus of rodents generate temporal sequential activity during sleep or immobility awake states. Some of these sequences have been shown to correlate with the previous (replay) or the future (preplay) spatial ordering of the corresponding place fields. Because of the abundant recurrent connections, these sequences are thought to originate in CA3 and trigger neuronal sequences downstream, e.g., in CA1. We previously showed that a continuous attractor model for the CA3 network can reliably generate long neuronal sequences (Azizi, Wiskott, & Cheng, 2013)⁠. These sequences were generated as a result of bump of activity moving through a preconfigured multi-chart network.

Experimental studies have shown that the chance of observing replay sequences is larger than that of the preplay sequences. Therefore our model should imprint the intrinsically generated sequences, such that noisy activity in the offline state have a higher chance of reactivating these imprinted sequences.

In this project, you will work on a previously implemented C-code of the network and implement an appropriate learning mechanism with the right parameters to model the replay phenomenon. Therefore, previous experience and a working knowledge of programming in C is required.

Literature:

Azizi, A. H., Wiskott, L., & Cheng, S. (2013). A computational model for preplay in the hippocampus. Frontiers in Computational Neuroscience, 7(August), 161. doi:10.3389/fncom.2013.00161

Supervisors:

 

Dr. Amir Azizi and Prof. Dr. Sen Cheng, Mercator Research Group "Structure of Memory", Faculty of Psychology

 

Back to top


Robust generation of spatio-temporal activity patterns in neuronal networks

Despite several decades of research the precise neuronal mechanisms underlying episodic memory, our memory of experienced events in our lives, remain unclear. We recently suggested that episodic memory are best represented as sequences of neural activity patterns unfolding in time and proposed specific contributions of the hippocampal subregions to the storage and retrieval of neuronal sequences (CRISP, see the literature). One central feature of CRISP is that hippocampal area CA3 intrinsically produces sequences. During memory encoding, intrinsic CA3 sequences are associated with sequences that are driven by sensory inputs. During memory retrieval, intrinsic CA3 sequences have to be reactivated based on partial, noisy cues. Therefore, the neural network mechanism in CA3 generating the sequences has to be robust to noise in the triggering cue. A number of neural networks have been proposed that can generate sequential activity, but their robustness to noise has rarely been studied. The goal of this project is to better understand the sensitivity of the various neural network models to noise. Since sequential neuronal activity is associated with a number of brain functions, e.g., movement, the results of this project are likely to be relevant far beyond the study of episodic memory. Prior programming experience is required.

Literature:

Cheng, S. (2013), The CRISP theory of hippocampal function in episodic memory. Frontiers in Neural Circuits, 7, 88. doi:10.3389/fncir.2013.00088

Supervisors:

Mehdi Bayati and Prof. Dr. Sen Cheng, Mercator Research Group "Structure of Memory", Faculty of Psychology

Back to top


 

Self-organization of spatially tuned neural activity in robotics simulations

Franzius et. al (2007) presented a model for the self-organized formation of place cells, head-direction cells, and spatial-view cells in the hippocampal formation based on unsupervised learning on quasi-natural visual stimuli. The model applies Slow Feature Analysis (SFA) to extract slowly varying features that are implicit in the visual input patterns. Which features emerge depends on the statistics of the movement of the simulated animal. For instance, if rotations are fast and translations are slow, the slowly varying features are related to spatial location. Likewise, if rotations are slow compared to translations, then the emerging features are related to head direction. These results were obtained by solving the SFA equations analytically and by computer simulations of virtual environments. The goal of this project is to replicate these results in robotics simulations in a physical environment. The task is far from trivial since physical constraints and noise introduce new challenges that have not been taken into account by the previous approaches. It is important to solve the challenges since the goal of the original study was to model the emergence of spatially tuned neural activity in biological brain that face similar challenges. A good command of Python is required.

Literature:

Franzius, M., Sprekeler, H., & Wiskott, L. (2007). Slowness and sparseness lead to place, head-direction, and spatial-view cells. PLoS Computational Biology, 3(8), e166. doi:10.1371/journal.pcbi.0030166

Supervisors:

Satish Kumar and Prof. Dr. Sen Cheng, Mercator Research Group "Structure of Memory", Faculty of Psychology

Back to top

 

 

Computational studies of the role of the semantic representation in episodic memory

Experimental studies have found that episodic memory in humans is unreliable, often preserving little more than the gist of the experienced episode, but none of the details. We hypothesize that this property of the episodic memory system results from the fact that episodes are stored only in terms of higher order information, i.e., semantic representation, not their underlying sensory inputs. This hypothesis is similar to Tulving's SPI (Serial-Parallel-Independent) model (Tulving, 1995), in which sensory information is first processed by the semantic system before being stored in the episodic system. We have recently developed a computational model to study the interrelation between the semantic and episodic system. In this model, the semantic system compresses the high-dimensional sensory inputs to a lower dimensionality both in space and time, i.e., the semantic representation can be represented by fewer neurons and varies at a lower rate than the sensory input stream. Episodic memories are represented as sequences of semantic representations, which are stored in and retrieved from a memory network. Preliminary results indicate that episodic memory is more accurate if the underlying semantic representation has been optimized for the object that is being stored in episodic memory. The goal of this project is to extend these results to new object types and classes of objects. A good command of Python is required.

Literature:

Tulving, E. (1995). Organization of memory: Quo vadis. In M. Gazzaniga (Ed.), The cognitive neurosciences (pp. 839–847). Cambridge, MA: MIT Press.

Supervisors:

Jing Fang and Prof. Dr. Sen Cheng, Mercator Research Group "Structure of Memory", Faculty of Psychology

Back to top

 

Necessary and sufficient behavioral evidence for episodic memory traces

Some of the most controversial topics in cognitive science are the question whether nonhuman animals possess a certain cognitive capacity or not. This is the case for episodic memory. Cheng et al. (2016) suggested that episodic memory can be dissociated into different components. Among them the episodic memory trace, semantic information and scenario generation. We argue that nonhuman animals probably share some, but perhaps not all, of the components of human episodic memory. In particular, we suggest that nonhuman animals can form and retrieve episodic memory traces, and thus remember the sequential order of events in an episode (Cheng & Werning, 2015). The goal of this project is to identify what behavioral evidence would be necessary and sufficient to claim that a nonhuman species possesses episodic memory traces. The project requires background knowledge of memory and experience in developing philosophical analyses.

Literature:

Cheng S, Werning M, Suddendorf T (2016) Dissociating memory traces and scenario construction in mental time travel. Neurosci Biobehav Rev 60:82–89.

Cheng S, Werning M (2015) What is Episodic Memory if it is a Natural Kind? Synthese.

Supervisors:

Prof. Dr. Sen Cheng, Institute for Neural Computation and Prof. Markus Werning, Institute for Philosophy II

Back to top

 

Modeling the dynamics of disease states in depression

Major depressive disorder (MDD) is a common and costly disorder associated with considerable morbidity, disability, and risk for suicide. The disorder is clinically and etiologically heterogeneous. Despite intense research efforts, the response rate of antidepressant treatments are relatively low and the etiology and progression of MDD remain poorly understood. We previously developed computational models to advance our understanding of MDD. First, we proposed a systematic and comprehensive definition of disease states, which is based on a type of mathematical model called a finite-state machine. Second, we proposed a dynamical systems model for the progression, or dynamics, of MDD and study under what conditions the model can account for the occurrence and recurrence of depressive episodes. The model is abstract, yet combines several major factors (mechanisms) that influence the dynamics of MDD. We used this model to study the effects of antidepressant treatments and cognitive behavioral therapy. The objective of this project is study whether different dynamics models can account better for epidemiological data than the previous model could. Prior programming experience is required.

Literature:

Demic S and Cheng S (2014) Modeling the dynamics of disease states in depression, PLoS ONE 9(10), e110358. doi: 10.1371/journal.pone.0110358

Supervisors:

Dr. Selver Demic and Prof. Dr. Sen Cheng, Mercator Research Group "Structure of Memory", Faculty of Psychology

Back to top

Processing of spatial information in the hippocampal circuit

Since grid cells were discovered in the medial entorhinal cortex, several models have been proposed for the transformation from periodic grids to the punctuate place fields of hippocampal place cells. For instance, Cheng and Frank identified the structure of the networks that is common to all robust solutions of the grids-to-places transformation [1]. Furthermore, we recently proposed that episodic memories, memories of personally experienced events, are best represented by temporal sequences of neural activation patterns (CRISP theory [2]). We have implemented CRISP in neural networks to store and retrieve episodic memories in a cortico-hippocampal network. One important feature of CRISP is that input sequences from grid cells are stored through synaptic plasticity in the feedforward projections of the hippocampus. In this project, we study whether the responses of neurons in the hippocampal sublayers (DG, CA3, and CA1) resemble the responses of recorded place cells. To this end, we let a virtual animal randomly explore an enclosure and record grid cells activity in medial entorhinal cortex based on the animal’s position as a sequence of activity patterns [3]. These sequences are then stored in the cortico-hippocampal network developed previously. Knowledge of the programming language Python is required.   

Literature:

[1] Cheng S and Frank LM. (2011), The structure of networks that produce the transformation from grid cells to place cells, Neuroscience, 197:293-306, doi: 10.1016/j.neuroscience.2011.09.002

[2] Cheng S (2013), The CRISP theory of hippocampal function in episodic memory, Front. Neural Circuits, 7:88, doi: 10.3389/fncir.2013.00088 

[3] Neher T, Cheng S and Wiskott L (2015), Memory storage fidelity in the hippocampal circuit: The role of subregions and input statistics. PLoS Comput Biol 11(5): e1004250., doi:10.1371/journal.pcbi.1004250

Supervisors:

Mehdi Bayati and Prof. Dr. Sen Cheng, Institute for Neural Computation, Mercator Research Group "Structure of Memory"

 

Back to top