May 12, 2022
ICASSP ‘22: Watch Dr Çağdaş Bilen’s industry expert talk on temporal decisions now
ICASSP 2022 began this week – it’s the world’s largest conference on acoustics, speech and signal processing – where, for a third year running we are presenting. Other speakers at the event include representatives from Amazon Lab 126, Meta, Sonos, IBM and many others.
Our Head of Research, Dr Çağdaş Bilen was presenting as part of the Industry Expert programme and his presentation is entitled ‘Closing the Gap Between Probabilities and Decisions in Temporal Detection and Classification’ and focuses on the importance of decision post-processing in temporal ML problems.
His talk went live today (Thursday 12th May) from 13:00 – 13:45 UTC. You can read more about the presentation on the ICASSP website here and watch the full presentation below:
You can download the corresponding poster here: https://www.audioanalytic.com/app/uploads/2022/04/ICASSP-2022-poster-Cagdas-Bilen-Audio-Analytic.pdf
Çağdaş will be available in the ICASSP GatherTown app during and after his keynote. If you are registered to attend this year’s ICASSP event, then you can come and meet us in Gather Area P.
Here is the synopsis for his talk:
Closing the Gap Between Probabilities and Decisions in Temporal Detection and Classification
Models are often designed and optimized to estimate instantaneous probabilities and rely on ‘ad-hoc decision post-processing’ to determine the occurrence of events. Within the constraints of commercial deployment, the impact of post-processing on the final performance of the product can be significant, sometimes reducing errors by orders of magnitude.
However, these constraints are often ignored in academic challenges and publications. Therefore the focus is mainly on improving the performance of the ML models. Furthermore, both in academia and industry, the design and optimization of the ML models often disregards the effect of such decision post-processing, hence potentially leading to suboptimal performance and products.