A present-day climatic model is presented in which extended wet spells of near-decadal duration and dry spells of similar length are explained on the basis of surface and upper tropospheric circulation variations. Wet speIls are shown to be the result of increased tropical atmospheric disturbances and tropical-temperate interaction, and to be linked to variations in the Walker Circulation. Conversely, dry spells are shown to result from diminished tropical activity over southern Africa, equatorward movement of westerly storm tracks and temperate perturbations in the westerlies.The present-day analogue is compared to preliminary spatial reconstructions of the climate of southern Africa over the last twenty-five millennia and is shown to have wide applicability in the explanation of the late-Quaternary palaeoelimates of the subcontinent. In particular, it is argued that the Last Glacial Maximum was associated with northward-displaced circulation conditions similar to those of present-day dry spells over the summer rainfall region, whereas the extensive moist conditions that prevailed for several thousand years after 9000 BP were analogous to present-day wet spell conditions with little apparent displacement of major circulation features.

We present a method for learning options from segmented demonstration trajectories. The trajectories are first segmented into skills using nonparametric Bayesian clustering and a reward function for each segment is then learned using inverse reinforcement learning. From this, a set of inferred trajectories for the demonstration are generated. Option initiation sets and termination conditions are learned from these trajectories using the one-class support vector machine clustering algorithm. We demonstrate our method in the four rooms domain, where an agent is able to autonomously discover usable options from human demonstration. Our results show that these inferred options can then be used to improve learning and planning.

No abstract