Increasing concentration of big data and computing resources has resulted in widespread adoption of machine learning as a service (MLaaS). The best-performing NLP, speech, image and video recognition tools are now provided as network services.
In such cases, the labelled data used for training may be proprietary, and different clients may be interested in different data distributions. This often violates the core ML generalizability assumption of the training and test distributions matching.
Join us in listening to Dr. Sunita Sarawagi this Saturday, November 27, 2021 at 5.pm as she discusses techniques to reduce such mismatch, and the ways in which servers can exploit multi-client training data to train ML models, for better generalization that suit client distributions.
She will also go into details of the methods that can be used for lightweight and heavyweight client adaptation of a blackbox service in the context of NLP models for topic adaptation, and speech models for accent adaptation.