Karsten Kube

Spike-timing-dependent plasticity in small-world networks

Biologically plausible excitatory neural networks develop a persistent synchronized pattern of activity depending on spontaneous activity and synaptic refractoriness (short term depression). By fixed synaptic weights synchronous bursts of oscillatory activity are stable and involve the whole network. In our modeling study we investigate the effect of a dynamic Hebbian-like learning mechanism, spike-timing-dependent plasticity (STDP), on the changes of synaptic weights depending on synchronous activity and network connection strategies (small-world topology). We show that STDP modifies the weights of synaptic connections in such a way that synchronization of neuronal activity is considerably weakened. Networks with a higher proportion of long connections can sustain a higher level of synchronization in spite of STDP influence. The resulting distribution of the synaptic weights in single neurons depends both on the global statistics of firing dynamics and on the number of incoming and outgoing connections.

Displaced strategies optimize connectivity in neocortical networks

This study considers the impact of different connection strategies in developing neocortical networks. An adequate connectivity is a requisite for synaptogenesis and the development of synchronous oscillatory network activity during maturation of cortical networks. In a defined time window early in development neurites have to grow out and connect to other neurons. Based on morphological observations we postulate that the underlying mechanism differs from common strategies of unspecific global or small-world strategies. We show that displaced connection strategies are very effective approaches to connect neurons with minimal wiring costs.

Contribution of the GABA shift to the transition from structural initialization to working stage in biologically realistic networks

Biological cortical neurons form functional networks through a complex set of developmental steps. A key process in early development is the transition of the spontaneous network dynamics from slow synchronous activity to a mature firing profile with complex high-order patterns of spikes and bursts. In the present modeling study we investigate the required properties of the network to initialize this transition by the shift of the chloride reversal potential, which switches the effect of the GABA synapses from depolarizing to hyperpolarizing. The simulated networks are generated by a statistical description of parameters for the neuron model and the network architecture.

Adaptation of Shape of Dendritic Spines by Genetic Algorithm

The role of dendritic spines in information processing of a neuron is still not clear. But it is known that they change their shape and size during learning processes. These effects may be important for storing of information (memory). We analyze the influence of shape variations on the electrical signal propagation in a group of dendritic spines by biologically realistic electrical simulation. In order to show the potential of shape changes a genetic algorithm is used to adapt the geometric parameters of the spine group to specific timing of incoming spikes. We can show that such a group of spines can do information processing like coincidence detection just by adjustment of its geometry.

Structural adaptation in young neocortical networks modeled by spatially coupled oscillators

The spontaneous synchronous activity in neocortical networks during early development is considered to be a requisite for the maturation of the networks. To analyze the structural adaptation of synaptic connections in large scale area, we simulate the network activity by distributed population models (complex oscillators). In this way we get a spatial distribution of parameters and activity and are able to study effects of local external stimulation and propagation of excitation waves in large network areas. Considering a small world connection strategy most connections are local but there are a number of long range connections, which work as shortcuts and help to synchronize the whole network activity. These long range connections can be used to adapt the network architecture by a Hebbian learning mechanism depending on the intrinsic wavelike network activity and external stimulation.

Influence of dendritic spines shape changes by learning

The role of dendritic spines in information processing of a neuron is still not clear. But it is known that they change its shape rapidly and fast during learning processes. These effects may be important for storing of information (memory). We show the influence of shape variations on the electrical signal propagation trough dendritic spines by biologically realistic electrical simulation. Basic properties of electrical signal transmission of single spines are estimated and approximated depend on the individual shape. Learning processes to adjust specific electrical properties are discussed and a possible mechanism is introduced.

Multi-population approach to approximate the development of neocortical networks

Cultured natural cortical neurons form functional networks through a complex set of developmental steps during the first weeks in vitro. The dynamic behavior of the network in this early development period changes from spontaneous spiking of single neurons to slow synchronous activity and finally to a mature firing profile with complex high-order patterns of spikes and bursts. In the present modeling study we investigate the required properties of the networks during the development by biologic realistic simulations and use an evolutionary algorithm (EA) to fit the parameters to the results of biological experiments. For each day in vitro (DIV) during the development a population of individuals is defined, which determines the statistical parameters to generate the networks and set up neuron properties by genes. The fitness function and the recombination algorithm are extended for this multi-population approach to allow the EA to follow different parameter trajectories over time (which are possible solutions) and include several kinds of biologically a-priori knowledge with an adjustable uncertainty.

Increased storage capacity in hopfield networks by small-world topology

We found via numerical simulations, that connectivity structure in sparsely connected hopfield networks for random bit patterns affects the storage capacity. Not only the number of local connections is important, but also, and in contrast, the recently found small-world-topology will increase the quality of recalled patterns. Here, we propose and investigate the impact from network network architecture to pattern storage capacity capabilities.

Liquid state machine by spatially coupled oscillators

Liquid State Machines [1] are a new strategie for real-time information processing in recurrent networks. In the present work we show that spatially coupled oscillators can be used as a usable liquid. If inputs stream are synchronized to oscillator phase its temporal dynamics can be be transformed into a high dimensional spatial pattern of oscillator activity. A memory less readout function can extract information about recent inputs. The fading memory is considered as the resynchronisation of oscillator field and can be adjusted by the parameter of small world connection mechanisms.

Simulation of young neocortical networks by spatially coupled oscillators

During early development neocortical neurons develop synchronous oscillatory network activity with the beginning of the second week in culture. This spontaneous synchronous activity is considered to be a requisite for the maturation of the synaptic networks. To simulate this activity it is common to use networks of neurons and analyze different connection strategies. On the other hand global population models (complex oscillators) are able to simulate the average activity of a large number of neurons. In the present work we combine both approaches to extend a population model to a set of distributed oscillators. In this way we get a spatial distribution of parameters and activity and are able to study effects of local external stimulation and propagation of excitation waves in large network areas. Small world connection strategy allows to investigate different behaviors by adjusting only one parameter.