Semih Akbayrak | BIASlab

Towards Universal Probabilistic Programming with Message Passing on Factor Graphs

Fri, 20 Jan 2023 00:00:00 +0000

Message Passing-based System Identification for NARMAX Models

Fri, 09 Dec 2022 09:00:00 +0200

Gaussian Process-based Amortization of Variational Message Passing Update Rules

Fri, 02 Sep 2022 00:00:00 +0000

Adaptive importance sampling message passing

Sun, 26 Jun 2022 00:00:00 +0000

Probabilistic programming with stochastic variational message passing

Wed, 22 Jun 2022 09:00:00 +0200

Adaptive Optimizer Design for Constrained Variational Inference

Wed, 01 Jun 2022 00:00:00 +0000

The Switching Hierarchical Gaussian Filter

Mon, 12 Jul 2021 00:00:00 +0000

Extended Variational Message Passing for Automated Approximate Bayesian Inference

Sat, 26 Jun 2021 16:37:31 +0200

Reparameterization Gradient Message Passing

Mon, 02 Sep 2019 09:00:00 +0200

CoHear

Wed, 01 Feb 2017 00:00:00 +0000

Hearing loss is a very serious health condition that has been associated with early dementia and cognitive decline. Still, hearing aid market penetration is quite low, in particular for the large population that is afflicted with ‘mild-to-moderate’ hearing impairment. This is mainly due to two reasons: Stigma (association with old age) and hearing aids (HA) sound quality. The recent commercial introduction of fashionable ‘hearables’ will likely alleviate the stigma issue. Recent advances in machine learning and cloud computing open new avenues for attacking the sound quality issue for hearing aids. In this project, we intend to develop a (crowd-based) collaborative design approach to improving the sound quality issues for the mild-to-moderately hearing-impaired population.

We expect to build a working prototype for collaboratively designed hearing algorithms that can be applied to a new class of ‘smart hearing devices’ with high appeal to the mild-to-moderately hearing-impaired patient. As an additional benefit, we hope that our technology will ease the transition from hearables to professional hearing aid technology for the moderate-to-profound hearing-impaired population.