Droughts and famines: The underlying factors and the causal links among agro-pastoral households in semi-arid Makueni district, Kenya

Famines are often linked to drought in semi-arid areas of Sub-Saharan Africa where not only pastoralists, but also increasingly agro-pastoralists are affected. This study addresses the interplay between drought and famine in the rural semi-arid areas of Makueni district, Kenya, by examining whether, and how crop production conditions and agro-pastoral strategies predispose smallholder households to drought-triggered food insecurity. If this hypothesis holds, then approaches to deal with drought and famine have to target factors causing household food insecurity during non-drought periods. Data from a longitudinal survey of 127 households, interviews, workshops, and daily rainfall records (1961–2003) were analysed using quantitative and qualitative methods. This integrated approach confirms the above hypothesis and reveals that factors other than rainfall, like asset and labour constraints, inadequate policy enforcement, as well as the poverty-driven inability to adopt risk-averse production systems play a key role. When linking these factors to the high rainfall variability, farmer-relevant definitions and forecasts of drought have to be applied.

Evaluating retraining rules for semi-supervised learning in neural network based cursive word recognition

Training a system to recognize handwritten words is a task that requires a large amount of data with their correct transcription. However, the creation of such a training set, including the generation of the ground truth, is tedious and costly. One way of reducing the high cost of labeled training data acquisition is to exploit unlabeled data, which can be gathered easily. Making use of both labeled and unlabeled data is known as semi-supervised learning. One of the most general versions of semi-supervised learning is self-training, where a recognizer iteratively retrains itself on its own output on new, unlabeled data. In this paper we propose to apply semi-supervised learning, and in particular self-training, to the problem of cursive, handwritten word recognition. The special focus of the paper is on retraining rules that define what data are actually being used in the retraining phase. In a series of experiments it is shown that the performance of a neural network based recognizer can be significantly improved through the use of unlabeled data and self-training if appropriate retraining rules are applied.