Using ideas from machine learning, Hsin-Hao has developed a more natural stimulus for stimulating visual cortical neurons. The simple white noise stimuli that are usually used are designed to be easy to analyse but do not strongly activate the neurons; natural stimuli activate the neurons better but are difficult to analyse. Hsin-Hao is developing an elegant way of transforming white noise to make it have more of the firing-inducing features of natural scenes but still be rigorously analysable.
With David MacKay, I have also used maximum-likelihood estimation to help find the membrane resistance and capacitance in a slice recording setup from the voltage response to a step of current. It is necessary to know the membrane resistance in order to determine the true membrane potential. Slight differences in the potential can make large differences in measured events. The membrane resistance is usually estimated by eye, but in difficult cases encountered in older mice slices, the automated approach is more accurate. The method also gives estimates of the cell membrane capacitance which is not easily obtained by eye.