DIKU Bits: Transfer learning to estimate influenza prevalence

Speaker

Ingemar J. Cox, professor in the Machine Learning Section at DIKU.

Abstract

It is now well-known that web searches can be used to estimate the prevalence of influenza (and other diseases) in a population. It is often claimed that this approach is especially useful in low and middle income countries (LMIC) where traditional health surveillance infrastructures are poor or absent. However, almost all proposed methods require training data that is often unavailable in LMIC regions. To address this issue, we describe a transfer learning solution where we train a model in say the USA, and then transfer the model to say Zimbabwe.

Zooming in on Ingemar Cox

Which courses do you teach at DIKU (BSc and MSc)?

Information retrieval, sometimes Web Science

Which technology/research/projects/start-ups are you exited to see develop?

AI for health

What is your favorite sketch from DIKUrevyen?

I don’t know what this is.