Correction: PPM-Decay: A computational model of auditory prediction with memory decay

AbstractStatistical learning and probabilistic prediction are fundamental processes in auditory cognition. A prominent computational model of these processes is Prediction by Partial Matching (PPM), a variable-order Markov model that learns by internalizing n-grams from training sequences. However, PPM has limitations as a cognitive model: in particular, it has a perfect memory that weights all historic observations equally, which is inconsistent with memory capacity constraints and recency effects observed in human cognition. We address these limitations with PPM-Decay, a new variant of PPM that introduces a customizable memory decay kernel. In three studies – one with artificially generated sequences, one with chord sequences from Western music, and one with new behavioral data from an auditory pattern detection experiment – we show how this decay kernel improves the model’s predictive performance for sequences whose underlying statistics change over time, and enables the model to capture effects of memory constraints on auditory pattern detection. The resulting model is available in our new open-source R package, ppm (https://github.com/pmcharrison/ppm).

Download Full-text

PPM-Decay: A computational model of auditory prediction with memory decay

PLoS Computational Biology ◽

10.1371/journal.pcbi.1008304 ◽

2020 ◽

Vol 16 (11) ◽

pp. e1008304

Author(s):

Peter M. C. Harrison ◽

Roberta Bianco ◽

Maria Chait ◽

Marcus T. Pearce

Keyword(s):

Computational Model ◽

Predictive Performance ◽

Pattern Detection ◽

Human Cognition ◽

Variable Order ◽

Probabilistic Prediction ◽

Auditory Pattern ◽

New Variant ◽

Memory Decay ◽

With Memory

Statistical learning and probabilistic prediction are fundamental processes in auditory cognition. A prominent computational model of these processes is Prediction by Partial Matching (PPM), a variable-order Markov model that learns by internalizing n-grams from training sequences. However, PPM has limitations as a cognitive model: in particular, it has a perfect memory that weights all historic observations equally, which is inconsistent with memory capacity constraints and recency effects observed in human cognition. We address these limitations with PPM-Decay, a new variant of PPM that introduces a customizable memory decay kernel. In three studies—one with artificially generated sequences, one with chord sequences from Western music, and one with new behavioral data from an auditory pattern detection experiment—we show how this decay kernel improves the model’s predictive performance for sequences whose underlying statistics change over time, and enables the model to capture effects of memory constraints on auditory pattern detection. The resulting model is available in our new open-source R package, ppm (https://github.com/pmcharrison/ppm).

Download Full-text

Self-Reported Sleep Relates to Microstructural Hippocampal Decline in β-Amyloid Positive Adults Beyond Genetic Risk

SLEEP ◽

10.1093/sleep/zsab110 ◽

2021 ◽

Author(s):

Håkon Grydeland ◽

Donatas Sederevičius ◽

Yunpeng Wang ◽

David Bartrés-Faz ◽

Lars Bertram ◽

...

Keyword(s):

Sleep Quality ◽

Genetic Risk ◽

Critical Role ◽

Sleep Efficiency ◽

Poor Sleep ◽

Time In Bed ◽

Polygenic Scores ◽

Memory Decay ◽

Β Amyloid ◽

With Memory

Abstract Study Objectives A critical role linking sleep with memory decay and β-amyloid (Aβ) accumulation, two markers of Alzheimer’s disease (AD) pathology, may be played by hippocampal integrity. We tested the hypotheses that worse self-reported sleep relates to decline in memory and intra-hippocampal microstructure, including in the presence of Aβ. Methods Two-hundred and forty-three cognitively healthy participants, aged 19-81 years, completed the Pittsburgh Sleep Quality Index once, and 2 diffusion tensor imaging sessions, on average 3 years apart, allowing measures of decline in intra-hippocampal microstructure as indexed by increased mean diffusivity. We measured memory decay at each imaging session using verbal delayed recall. One session of positron emission tomography, in 108 participants above 44 years of age, yielded 23 Aβ positive. Genotyping enabled control for APOE ε4 status, and polygenic scores for sleep and AD, respectively. Results Worse global sleep quality and sleep efficiency related to more rapid reduction of hippocampal microstructure over time. Focusing on efficiency (the percentage of time in bed at night spent asleep), the relation was stronger in presence of Aβ accumulation, and hippocampal integrity decline mediated the relation with memory decay. The results were not explained by genetic risk for sleep efficiency or AD. Conclusions Worse sleep efficiency related to decline in hippocampal microstructure, especially in the presence of Aβ accumulation, and Aβ might link poor sleep and memory decay. As genetic risk did not account for the associations, poor sleep efficiency might constitute a risk marker for AD, although the driving causal mechanisms remain unknown.

Download Full-text