Predictive coding and stochastic resonance as fundamental principles of auditory perception

Abstract

Cognitive computational neuroscience (CCN) suggests that to gain a mechanistic understanding of brain function, hypothesis driven experiments should be accompanied by biologically plausible computational models. This novel research paradigm offers a way from alchemy to chemistry, in auditory neuroscience. With a special focus on tinnitus -as the prime example of auditory phantom perception -we review recent work at the intersection of artificial intelligence, psychology, and neuroscience, foregrounding the idea that experiments will yield mechanistic insight only when employed to test formal or computational models. This view challenges the popular notion that tinnitus research is primarily data limited, and that producing large, multi-modal, and complex data-sets, analyzed with advanced data analysis algorithms, will lead to fundamental insights into how tinnitus emerges. We conclude that two fundamental processing principles -being ubiquitous in the brain -best fit to a vast number of experimental results and therefore provide the most explanatory power: predictive coding as a top-down, and stochastic resonance as a complementary bottom-up mechanism. Furthermore, we argue that even though contemporary artificial intelligence and machine learning approaches largely lack biological plausibility, the models to be constructed will have to draw on concepts from these fields; since they provide a formal account of the requisite computations that underlie brain function. Nevertheless, biological fidelity will have to be addressed, allowing for testing possible treatment strategies in silico, before application in animal or patient studies. This iteration of computational and empirical studies may help to open the 'black boxes' of both machine learning and the human brain.

Intrinsic Noise Improves Speech Recognition in a Computational Model of the Auditory Pathway

“…Furthermore, our model is [like the central gain ( Auerbach et al, 2014 ) and the lateral inhibition model ( Gerken, 1996 )] a pure bottom-up model, which means that cortical or thalamocortical top-down modulations are not regarded. Note that, we do not discuss further bottom-up models of tinnitus development in detail, as these models make no predictions on speech perception benefit of tinnitus after hearing loss (for an in-depth comparison of the different models, see Schilling et al, 2022 ). In contrast to bottom-up models, top- down models play a crucial role in understanding why brainstem hyperactivity passes the “gate to consciousness” (the thalamus) and results in suffering a psychic burden.…”

“…More recent approaches suggest that tinnitus is a prediction error and formalize their models within the Bayesian brain framework ( Sedley et al, 2016 ; Hullfish et al, 2019 ; De Ridder and Vanneste, 2021 ). In summary, it is necessary to merge bottom-up and top-down models of tinnitus development to achieve a unified explanation of tinnitus development ( Schilling et al, 2022 ). Our bottom-up model has not exclusively explanatory power but might also serve as source of inspiration for advanced machine learning approaches.…”

“…Remarkably, a cerebellar-like neuronal architecture resembling such delay-lines is known to exist in the DCN ( Osen et al, 1988 ; Hackney et al, 1990 ; Nelken and Young, 1994 ; Oertel and Young, 2004 ; Baizer et al, 2012 ). Therefore, we previously proposed the possibility that the neural noise for SR is injected into the auditory system via somatosensory projections to the DCN ( Krauss et al, 2016 , 2018 , 2019b ; Krauss and Tziridis, 2021 ; Schilling et al, 2021d , 2022 ). The idea that central noise plays a key role in auditory processing has recently gained increasing popularity ( Zeng, 2013 , 2020 ; Koops and Eggermont, 2021 ) and is supported by various findings.…”

“…The aim of our implementation is not to understand the whole auditory pathway in detail, which would be far to ambitious, but to find out if SR could have a significant effect on speech perception. Thus, it is not the aim of the study to analyze the auditory system on an implementational level (see Marrs’ level of analysis; Marr and Poggio, 1979 ), but to explain the algorithmic level ( Krauss and Schilling, 2020 ; Schilling et al, 2022 ).…”

Deep learning based decoding of local field potential events

“…However, our findings that lag-times correspond to different LFP event shapes and that lag-time histograms show two maxima indicate that standard synchrony analyses fail to provide the full picture. Applying the here presented novel approach to further investigate the functional plasticity after hearing loss, which is hypothesized to be the cause of tinnitus [60][61][62][63][64][65][66], might lead to a deeper understanding of the underlying processes.…”