Unsupervised clustering of temporal patterns in high-dimensional neuronal ensembles using a novel dissimilarity measure

Abstract

Temporally ordered multi-neuron patterns likely encode information in the brain. We introduce an unsupervised method, SPOTDisClust (Spike Pattern Optimal Transport Dissimilarity Clustering), for their detection from high-dimensional neural ensembles. SPOTDisClust measures similarity between two ensemble spike patterns by determining the minimum transport cost of transforming their corresponding normalized cross-correlation matrices into each other (SPOTDis). Then, it performs density-based clustering based on the resulting inter-pattern dissimilarity matrix. SPOTDisClust does not require binning and can detect complex patterns (beyond sequential activation) even when high levels of out-of-pattern “noise” spiking are present. Our method handles efficiently the additional information from increasingly large neuronal ensembles and can detect a number of patterns that far exceeds the number of recorded neurons. In an application to neural ensemble data from macaque monkey V1 cortex, SPOTDisClust can identify different moving stimulus directions on the sole basis of temporal spiking patterns.

Entorhinal-retrosplenial circuits for allocentric-egocentric transformation of boundary coding

“…Typical border cell classification using the original border score ( Solstad et al, 2008 ) identified only a small fraction of border cells in RSC, as this score is based on the occupancy of a single firing field along a wall and is strongly biased to connected bins ( Figure 1—figure supplement 2A–C ). We thus developed a new model-based approach using a template-matching procedure to classify these border cells in RSC ( Figure 1D–F ), based on Grossberger et al, 2018 .…”

“…Several additional templates were constructed to assess the effects of behavioral manipulation, adding additional weight in the location of placed objects/walls ( Figure 2E,J ). The EMD distance between a rate map and a template represents the minimal cost that must be paid to transform one distribution into another, with values ranging between zero (identical maps) and one (maximal difference), and is thus a normalized metric of dissimilarity ( Grossberger et al, 2018 ).…”

Bayesian inference of neuronal assemblies

“…Therefore, the functional segregation of subnetworks of assemblies may be relevant for understanding how activation of different tectal locations leads to mutually exclusive behaviors [30–32]. The assembly activity matrix, obtained from our method, can also be used to identify temporal patterns of assembly activation sequences [33] that might be more relevant than single assembly firing events for understanding tectally-mediated behavior.…”

Classification of Agricultural Emissions Among OECD Countries with Unsupervised Techniques

“…HDBSCAN method was used by Rahman (2017) to augment location based services on the Internet (Rahman et al, 2016). It was also used in many biological applications, for instance in clustering of temporal patterns in highdimensional neuronal ensembles (Grossberger, 2018).…”