Search

Multiplicative Weights Update (MWU) algorithms are a form of online learning that is applied to multi-armed bandit problems. Such problems involve allocating a fixed number of trials among multiple options to maximize cumulative payoff. MWU is a popular and effective method for dynamically balancing the trade-off between exploring the value of new options and exploiting the information already gained. However, no clear strategy exists to help practitioners choose which of the several algorithmic designs within this family to deploy. In this paper, three variants of parallel MWU algorithms are considered: Two parallel variants that rely on global memory, and one variant that uses distributed memory. The three variants are first analyzed theoretically, and then their effectiveness is assessed empirically on the task of estimating distributions in the context of stochastic search for repairs to bugs in software. Earlier work on APR suffers from various inefficiencies, and the paper shows how to decompose the problem into two stages: one that is embarrassingly parallel and one that is amenable to MWU. We then model the cost of each MWU variant and derive the conditions under which it is likely to be preferred in practice. We find that all three MWU algorithms achieve accuracy above 90% but that there are significant differences in runtime and total cost. When 90% accuracy is sufficient and evaluating options is expensive, such as in our use case, we find that the algorithm that uses global memory and has high communication cost outperforms the other two. We analyze the reasons for this surprising result.

A number of satellite missions have the capability to provide surface soil moisture information at a range of spatial and temporal scales. However, the validation of such products heavily relies on point measurements from permanent stations, which may or may not be representative of the larger scale soil moisture conditions. Hence, methods need to be developed that allow the sampling of surface soil moisture on the ground across large scales over a reasonably short time scale, in order to capture the spatial variability within a footprint, or to provide spatially sufficiently large data sets to validate high-resolution products, be they at their native resolution or downscaled. In this study, two field-scale ground sampling techniques, namely stationary and roving Cosmic Ray Probes, are compared against a high-resolution satellite product. The data are compared for their temporal performance as well as with a focus on capturing the correct spatial variability. The challenge is the inherently different sensing depth of the various technologies. It is shown that this may largely be overcome through scaling the products.

The L-band (1 – 2 GHz) microwave remote sensing has been widely acknowledged as the most promising method to monitor regional to global soil moisture. Consequently, the Soil Moisture Active Passive (SMAP) satellite applied this technique to provide global soil moisture every 2 to 3 days. To verify the performance of SMAP, the fourth and fifth campaign of SMAP Experiments (SMAPEx-4 & -5) were carried out at the beginning of the SMAP operational phase in the Murrumbidgee River catchment, southeast Australia. The airborne radar and radiometer observations together with ground sampling on soil moisture, vegetation water content, and surface roughness were collected in coincidence with SMAP overpasses. The SMAPEx-4 & -5 data sets will benefit to SMAP post-launch calibration and validation under Australian land surface conditions.

• Drought has many faces, monitoring them requires a vast range of observations • Key questions: who is affected, how bad is it, who's fault is it, and will it get better? • Recent legislation has made the Bureau of Meteorology the natural home for drought monitoring systems. • The Australian Water Resources Assessment (AWRA) system is currently in the operational testing phase • Earth observation plays a critical role where the on-ground network is lacking • Chestnut 1: detail and accuracy vs. cost and usefulness • Chestnut 2: Australia does not have space agency or program.

Many multitemporal change detection studies require calibrated sequences of satellite imagery to monitor changes in the Earth surface free of the affects of atmospheric and other influences. Two image calibration procedures were assessed for their ability to estimate the atmospheric contribution to the satellite signal from the calibration results. A sequence of TM and ETM+ imagery over Hay (93/84) was used in the analysis because of the availability of the necessary in situ data. The assessment focussed on the comparison of modeled (from in situ data) and estimated (from the calibration results) values of digital counts associated with a zero surface reflectance. Generally, best agreement with modeled values was observed for band 1-3. Bands 4-7 typically displayed poorer agreement with modeled values due mainly to the increased scatter about the regression lines.