The published paper was presented on the 17th of November at ACM SenSys AIChallengeIoT 2021.
PhD researcher from TU Delft, one of the partners of SPATIAL and coordinator of the project, is leading this work, in collaboration with Nokia Bell Labs.
This published work explores the AI fairness angle in audio-based embedded machine learning. It advocates a data-centric approach and illustrates how pre-processing parameters can trade-off between a model’s accuracy, fairness, and system efficiency, as well as to make an embedded ML model resilient to unseen deployment conditions.
When deploying machine learning (ML) models on embedded and IoT devices, performance encompasses more than an accuracy metric: inference latency, energy consumption, and model fairness are necessary to ensure reliable performance under heterogeneous and resource-constrained operating conditions.
To this end, prior research has studied model-centric approaches, such as tuning the hyperparameters of the model during training and later applying model compression techniques to tailor the model to the resource needs of an embedded device.
In this paper, the data-centric view of embedded ML is taken, and study the role that pre-processing parameters in the data pipeline can play in balancing the various performance metrics of an embedded ML system. Through an in-depth case study with audio-based keyword spotting (KWS) models, we show that pre-processing parameter tuning is a remarkable tool that model developers can adopt to trade-off between a model’s accuracy, fairness and system efficiency, as well as to make an embedded ML model resilient to unseen deployment conditions.
The paper is open access in the ACM digital library – SPATIAL project is acknowledged:
For further information: https://dl.acm.org/doi/10.1145/3485730.3493448