In this paper, we propose conjugate energy-based models (CEBMs), a new class of energy-based models that define a joint density over data and latent variables. The joint density of a CEBM decomposes into an intractable distribution over data and a tractable posterior over latent variables. CEBMs have similar use cases as variational autoencoders, in the sense that they learn an unsupervised mapping from data to latent variables. However, these models omit a generator network, which allows them to learn more flexible notions of similarity between data points. Our experiments demonstrate that conjugate EBMs achieve competitive results in terms of image modelling, predictive power of latent space, and out-of-domain detection on a variety of datasets.

Last week, a student asked me whether our new MSc module "Big Data Epidemiology" would be covering "machine learning" techniques and enthusiastically told me all about how they intend to apply such techniques to their own research. The short answer to the student's question was "Yes, but only briefly". The long answer requires some exploration into what we mean by "big data epidemiology" and consideration of what machine learning can (and perhaps more importantly, cannot) do for researchers. Data is often considered "big data" if it can be described in terms of the "four V's": volume (there's a lot of it), variety (the data takes lots of different forms), velocity (the data changes or is updated frequently) and veracity (the data may be of poor/ unknown quality).1 On our module we use electronic healthcare records databases that can contain information about millions of patients, collected over several years (volume).