London School of Economics and Political Science

About: London School of Economics and Political Science is a education organization based out in London, United Kingdom. It is known for research contribution in the topics: Politics & Population. The organization has 8759 authors who have published 35017 publications receiving 1436302 citations.

PRIME: A Programme to Reduce the Treatment Gap for Mental Disorders in Five Low- and Middle-Income Countries

Abstract: Crick Lund and colleagues describe their plans for the PRogramme for Improving Mental health carE (PRIME), which aims to generate evidence on implementing and scaling up integrated packages of care for priority mental disorders in primary and maternal health care contexts in Ethiopia, India, Nepal, South Africa, and Uganda.

QSGD: Communication-Efficient SGD via Gradient Quantization and Encoding

Abstract: Parallel implementations of stochastic gradient descent (SGD) have received significant research attention, thanks to excellent scalability properties of this algorithm, and to its efficiency in the context of training deep neural networks. A fundamental barrier for parallelizing large-scale SGD is the fact that the cost of communicating the gradient updates between nodes can be very large. Consequently, lossy compression heuristics have been proposed, by which nodes only communicate quantized gradients. Although effective in practice, these heuristics do not always provably converge, and it is not clear whether they are optimal. In this paper, we propose Quantized SGD (QSGD), a family of compression schemes which allow the compression of gradient updates at each node, while guaranteeing convergence under standard assumptions. QSGD allows the user to trade off compression and convergence time: it can communicate a sublinear number of bits per iteration in the model dimension, and can achieve asymptotically optimal communication cost. We complement our theoretical results with empirical data, showing that QSGD can significantly reduce communication cost, while being competitive with standard uncompressed techniques on a variety of real tasks. In particular, experiments show that gradient quantization applied to training of deep neural networks for image classification and automated speech recognition can lead to significant reductions in communication cost, and end-to-end training time. For instance, on 16 GPUs, we are able to train a ResNet-152 network on ImageNet 1.8x faster to full accuracy. Of note, we show that there exist generic parameter settings under which all known network architectures preserve or slightly improve their full accuracy when using quantization.

On the Speed of Transition Central Europe

Abstract: Transition in Central Europe is four years old. State firms which dominated the economy are struggling with market forces. A new private sector quickly emerged and has taken hold. Unemployment, which did not exist, is high and still increasing. Will this process of transition accelerate, or slow down? Will unemployment keep increasing? Can things go wrong and how? Our paper represents a first pass at answering those questions. The basic structure of the model we develop is standard, that of the transition from a low to a high productivity sector. But we pay attention to two aspects which strike us as important. The first is the interactions between unemployment and the decisions of both state and private firms. The second are the idiosyncracies which come from the central planning legacy, from the structure of control within state firms to the lack of many market institutions, which limits private sector growth. We start with a description of transition in Poland so far. We then develop a model and use it to think about the determinants of the speed of transition and the level of unemployment. Finally, we return to the role of policy and the future in Poland, as well as the causes of cross-Central European country variations.