## sorting: Bayesian clustering of nonstationary data (2004)

Venue: | In: Proceedings of the 18th International Conference on Neural Information Processing Systems |

Citations: | 14 - 1 self |

### BibTeX

@INPROCEEDINGS{Bar-hillel04sorting:bayesian,

author = {Aharon Bar-hillel and Adam Spiro and Eran Stark},

title = {sorting: Bayesian clustering of nonstationary data},

booktitle = {In: Proceedings of the 18th International Conference on Neural Information Processing Systems},

year = {2004}

}

### OpenURL

### Abstract

Spike sorting involves clustering spike trains recorded by a microelectrode according to the source neuron. It is a complicated problem, which requires a lot of human labor, partly due to the non-stationary nature of the data. We propose an automated technique for the clustering of non-stationary Gaussian sources in a Bayesian framework. At a first search stage, data is divided into short time frames and candidate descriptions of the data as a mixture of Gaussians are computed for each frame. At a second stage transition probabilities between candidate mixtures are computed, and a globally optimal clustering is found as the MAP solution of the resulting probabilistic model. Transition probabilities are computed using local stationarity assumptions and are based on a Gaussian version of the Jensen-Shannon divergence. The method was applied to several recordings. The performance appeared almost indistinguishable from humans in a wide range of scenarios, including movement, merges, and splits of clusters. 1

### Citations

9231 |
Elements of Information Theory
- Cover, Thomas
- 1990
(Show Context)
Citation Context ...n two consecutive time frames we saw two labeled samples (X1, L1), (X2, L2) of sizes N1, N2 with empirical distributions Θ1, Θ2 respectively. By ’empirical distribution’, or ’type’ in the notation of =-=[2]-=-, we denote the ML parameters of the sample, for both the multinomial distribution of the mixture weights and the Gaussian distributions of the components. As stated above, we assume that the joint sa... |

547 |
Testing Statistical Hypotheses
- Lehmann
- 1986
(Show Context)
Citation Context ...ic family, the JS-divergence is a non-negative measurement which can be used to test whether several samples are derived from a single distribution from the family or from a mixture of different ones =-=[6]-=-. The JS-divergence is computed for a mixture of n empirical distributions P1, .., Pn with mixture weights π1, .., πn. In the Gaussian case, denote the mean and covariance of the component distributio... |

138 |
A review of methods for spike sorting: The detection and classification of neural action potentials
- Lewicki
- 1998
(Show Context)
Citation Context ...s task is mostly done manually. It is a tedious mission, requiring many hours of human work for a single recording session. Many algorithms were proposed in order to help automating this process (see =-=[7]-=- for a review, [9],[10]), and some were implemented to provide a helping tool for manual sorting [8]. However, the ability of suggested algorithms to replace the human worker has been quite limited. O... |

116 |
The Hungarian method for the assignment problem. Naval Research Logistics Quarterly 2(1):83–97
- Kuhn
- 1955
(Show Context)
Citation Context ...es of the graph are the Gaussian components of Θ1, Θ2 and the edges’ weights are given by the scores S(a, b). The global optimum of this problem can be efficiently found using the Hungarian algorithm =-=[5]-=- in O(n 3 ), which is unproblematic in our case. The one-to-one correspondence assumption is too strong for many data sets in the spike sorting application, as it ignores the phenomena of splits and m... |

50 |
Multispike train analysis
- Abeles, MH
- 1977
(Show Context)
Citation Context ...em, since electrodes are placed in a single location for long durations. Using a representation with the first 2 PCA coefficients (which preserves up to 93% of the variance of the original recordings =-=[1]-=-), a human can cluster spikes by visual inspection. When dividing the data into small enough time frames, a cluster density can be approximated by a multivariate Gaussian with a general covariance mat... |

32 |
Erratum: Automatic sorting of multiple unit neuronal signals in the presence of anisotropic and non-gaussian variability (j. neurosci. methods 69
- Fee, Mitra, et al.
- 1996
(Show Context)
Citation Context .... A split of a single cluster is possible in the same manner. Most spike sorting algorithms don’t address the presented difficulties at all, as they assume full stationarity of the data. Some methods =-=[4, 11]-=- try to cope with the lack of stationarity by grouping data into many small clusters and identifying the clusters that can be combined to represent the activity of a single unit. In the second stage, ... |

26 |
Robust, automatic spike sorting using mixtures of multivariate t-Distribution
- Shoham, Fellow, et al.
- 2003
(Show Context)
Citation Context ...manually. It is a tedious mission, requiring many hours of human work for a single recording session. Many algorithms were proposed in order to help automating this process (see [7] for a review, [9],=-=[10]-=-), and some were implemented to provide a helping tool for manual sorting [8]. However, the ability of suggested algorithms to replace the human worker has been quite limited. One of the main obstacle... |

8 |
Classification of nonstationary neural signals
- Snider, Bonds
- 1998
(Show Context)
Citation Context .... A split of a single cluster is possible in the same manner. Most spike sorting algorithms don’t address the presented difficulties at all, as they assume full stationarity of the data. Some methods =-=[4, 11]-=- try to cope with the lack of stationarity by grouping data into many small clusters and identifying the clusters that can be combined to represent the activity of a single unit. In the second stage, ... |

2 |
Rebrik S.P, Kurgansky A.V, Miller K.D. Tracking neurons recorded from tetrodes across time
- Emondi
- 2004
(Show Context)
Citation Context ... activity of a single unit. In the second stage, [4] uses ISI information to understand which clusters cannot be combined, while [11] bases this decision on the density of points between clusters. In =-=[3]-=- a semi-automated method is suggested, in which each time frame is clustered manually, and then the correspondence between clusters in consecutive time frames is established automatically. The corresp... |

2 | Characterization of and compensation for the nonstationarity of spike shapes during physiological recordings. Neurocomputing
- Penev, Dimitrov, et al.
- 2001
(Show Context)
Citation Context ...one manually. It is a tedious mission, requiring many hours of human work for a single recording session. Many algorithms were proposed in order to help automating this process (see [7] for a review, =-=[9]-=-,[10]), and some were implemented to provide a helping tool for manual sorting [8]. However, the ability of suggested algorithms to replace the human worker has been quite limited. One of the main obs... |