This paper studies three related algorithms: the (traditional) Gradient Descent (GD) Algorithm, the Exponentiated Gradient Algorithm with Positive and Negative weights (EG ¡ algorithm) and the Exponentiated Gradient Algorithm with Unnormalized Positive and Negative weights (EGU ¡ algorithm). These algorithms have been previously analyzed using the "mistake-bound framework" in the computational learning theory community. In this paper we perform a traditional signal processing analysis in terms of the mean square error. A relationship between the learning rate and the mean squared error (MSE) of predictions is found for the family of algorithms. This is used to compare the performance of the algorithms by choosing learning rates such that they converge to the same steady state MSE. We demonstrate that if the target weight vector is sparse, the EG ¡ algorithm typically converges more quickly than the GD or EGU ¡ algorithms which perform very similarly. A side effect of our analysis is a reparametrization of the algorithms that provides insights into their behavior. The general form of the results we obtain are consistent with those obtained in the mistake-bound framework [1]. The application of the algorithms to acoustic echo cancellation is then studied and it is shown in some circumstances that the EG ¡ algorithm will converge faster than the other two algorithms.

This paper attempts to distil the key conclusions from the very large literature on the empirics of growth and to apply them to the development record of the five major Southeast Asian economies for which we have reasonably long-term data — Indonesia, Malaysia, The Philippines, Singapore and Thailand. These five display a considerable range of development outcomes, ranging from consistently high growth, to episodes of boom and crisis, and to low average growth. After estimating a series of general empirical models from a large sample of countries, we examine how well these fit the observed outcomes in these particular Southeast Asian countries. Our broad finding is that the average model does reasonably well in explaining outcomes in Singapore and Thailand, but that the residuals for Indonesia, Malaysia and the Philippines are quite large and persistent across different specifications.

No abstract

No abstract

JT03305647 Improving Access and Quality in the Indian Education SystemEducation has been given high priority by India's central and state governments and continues to grow fast. School access has been expanded by investment in school infrastructure and recruitment of teachers. In higher education too, the number of providers continues to rise rapidly. A new law enshrining the rights of all children to free and compulsory education will further lift enrolment, bringing closer the government's goal of universal elementary education, which comprises eight years of schooling. Nevertheless, high drop-out rates and low attendance continues to be a challenge at lower levels and enrolment at higher levels remains modest by international standards. Private sector involvement is on the rise. While it helps expand education infrastructure, particularly in higher education, access has not always been assured and the availability of student loans for higher education needs to improve. Poor learning outcomes amongst school students and mediocre higher education provision call for more effective government regulation and funding arrangements. Expanding resources will help but they need to be deployed more effectively, while incentives and professional development systems for teachers need to be strengthened. In higher education the government has proposed reforms which have the potential to bring about much-needed improvements in regulatory effectiveness. Efforts should focus on reducing micro-regulation and improving institutional autonomy, in order to stimulate innovation and diversity. Increasing the number of institutions subjected to quality assessments will be important for lifting standards across the higher education system, while reform of recruitment and promotion mechanisms could help attract and retain talent in academia.