Dataset Information

Bayesian online learning of the hazard rate in change-point problems.

ABSTRACT: Change-point models are generative models of time-varying data in which the underlying generative parameters undergo discontinuous changes at different points in time known as change points. Change-points often represent important events in the underlying processes, like a change in brain state reflected in EEG data or a change in the value of a company reflected in its stock price. However, change-points can be difficult to identify in noisy data streams. Previous attempts to identify change-points online using Bayesian inference relied on specifying in advance the rate at which they occur, called the hazard rate (h). This approach leads to predictions that can depend strongly on the choice of h and is unable to deal optimally with systems in which h is not constant in time. In this letter, we overcome these limitations by developing a hierarchical extension to earlier models. This approach allows h itself to be inferred from the data, which in turn helps to identify when change-points occur. We show that our approach can effectively identify change-points in both toy and real data sets with complex hazard rates and how it can be used as an ideal-observer model for human and animal behavior when faced with rapidly changing inputs.

SUBMITTER: Wilson RC

PROVIDER: S-EPMC2966286 | biostudies-literature | 2010 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Bayesian online learning of the hazard rate in change-point problems.

Wilson Robert C RC Nassar Matthew R MR Gold Joshua I JI

Neural computation 20100901 9

Change-point models are generative models of time-varying data in which the underlying generative parameters undergo discontinuous changes at different points in time known as change points. Change-points often represent important events in the underlying processes, like a change in brain state reflected in EEG data or a change in the value of a company reflected in its stock price. However, change-points can be difficult to identify in noisy data streams. Previous attempts to identify change-po ...[more]

PMID: 20569174

Similar Datasets

Project description:Reaction time testing is widely used in online computerized concussion assessments, and most concussion studies utilizing the metric have demonstrated varying degrees of difference between concussed and non-concussed individuals. The problem with most of these online concussion assessments is that they predominantly rely on consumer grade technology. Typical administration of these reaction time tests involves presenting a visual stimulus on a computer monitor and prompting the test subject to respond as quickly as possible via keypad or computer mouse. However, inherent delays and variabilities are introduced to the reaction time measure by both computer and associated operating systems that the concussion assessment tool is installed on. The authors hypothesized systems that are typically used to collect concussion reaction time data would demonstrate significant errors in reaction time measurements. To remove human bias, a series of experiments was conducted robotically to assess timing errors introduced by reaction time tests under four different conditions. In the first condition, a visual reaction time test was conducted by flashing a visual stimulus on a computer monitor. Detection was via photodiode and mechanical response was delivered via computer mouse. The second condition employed a mobile device for the visual stimulus, and the mechanical response was delivered to the mobile device's touchscreen. The third condition simulated a tactile reaction time test, and mechanical response was delivered via computer mouse. The fourth condition also simulated a tactile reaction time test, but response was delivered to a dedicated device designed to store the interval between stimulus delivery and response, thus bypassing any problems hypothesized to be introduced by computer and/or computer software. There were significant differences in the range of responses recorded from the four different conditions with the reaction time collected from visual stimulus on a mobile device being the worst and the device with dedicated hardware designed for the task being the best. The results suggest that some of the commonly used visual tasks on consumer grade computers could be (and have been) introducing significant errors for reaction time testing and that dedicated hardware designed for the reaction time task is needed to minimize testing errors.

Dataset Information

Bayesian online learning of the hazard rate in change-point problems.

Publications

Bayesian online learning of the hazard rate in change-point problems.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets