Earthquakes, influenza and cycles of Indian kala-azar.

It is suggested that previous data indicate 3 major epidemics of kala-azar in Assam between 1875 and 1950, with inter-epidemic periods of 30-45 and 20 years. This deviates from the popular view of regular cycles with a 10-20 year period. A deterministic mathematical model of kala-azar is used to find the simplest explanation for the timing of the 3 epidemics, paying particular attention to the role of extrinsic (drugs, natural disasters, other infectious diseases) versus intrinsic (host and vector dynamics, birth and death rates, immunity) processes in provoking the second. We conclude that, whilst widespread influenza in 1918-1919 may have magnified the second epidemic, intrinsic population processes provide the simplest explanation for its timing and synchrony throughout Assam. The model also shows that the second inter-epidemic period is expected to be shorter than the first, even in the absence of extrinsic agents, and highlights the importance of a small fraction of patients becoming chronically infectious (with post kala-azar dermal leishmaniasis) after treatment during an epidemic.

Scaling the Indian Buffet Process via Submodular Maximization

Inference for latent feature models is inherently difficult as the inference space grows exponentially with the size of the input data and number of latent features. In this work, we use Kurihara & Welling (2008)'s maximization-expectation framework to perform approximate MAP inference for linear-Gaussian latent feature models with an Indian Buffet Process (IBP) prior. This formulation yields a submodular function of the features that corresponds to a lower bound on the model evidence. By adding a constant to this function, we obtain a nonnegative submodular function that can be maximized via a greedy algorithm that obtains at least a one-third approximation to the optimal solution. Our inference method scales linearly with the size of the input data, and we show the efficacy of our method on the largest datasets currently analyzed using an IBP model.