Saikat Sinha Roy

Handwritten isolated Bangla compound character recognition: A new benchmark using a novel deep learning approach

Abstract In this work, a novel deep learning technique for the recognition of handwritten Bangla isolated compound character is presented and a new benchmarkof recognition accuracy on the CMATERdb 3.1.3.3 dataset is reported. Greedy layer wise training of Deep Neural Network has helped to make significant strides in various pattern recognition problems. We employ layerwise training to Deep Convolutional Neural Networks (DCNN) in a supervised fashion and augment the training process with the RMSProp algorithm to achieve faster convergence. We compare results with those obtained from standard shallow learning methods with predefined features, as well as standard DCNNs. Supervised layerwise trained DCNNs are found to outperform standard shallow learning models such as Support Vector Machines as well as regular DCNNs of similar architecture by achieving error rate of 9.67% thereby setting a new benchmark on the CMATERdb 3.1.3.3 with recognition accuracy of 90.33%, representing an improvement of nearly 10%.

Convolutional regression framework for health behavior prediction

Understanding the propagation of human health behavior, such as smoking and obesity, and identification of the factors that control such phenomenon is an important area of research in recent years mainly because, in industrialized countries a substantial proportion of the mortality and quality of life is due to particular behavior patterns, and that these behavior patterns are modifiable. Predicting the individuals who are going to be overweight or obese in future, as overweight and obesity propagate over dynamic human interaction network, is an important problem in this area. However, the problem has received limited attention from the network analysis and machine learning perspective till date. In this work, we propose a scalable supervised prediction model based on convolutional regression framework that is particularly suitable for short time series data. We propose various schemes to model social influence for health behavior change. Further we study the contribution of the primary factors of overweight and obesity, like unhealthy diets, recent weight gains and inactivity in the prediction task. A thorough experiment shows the superiority of the proposed method over the state-of-the-art.

Foreign Direct Investment, Firm Heterogeneity, and Exports: An Analysis of Indian Manufacturing

Using firm-level data, this paper investigates whether foreign direct investment and the presence of multinational enterprises explains Indias improved export performance during the postreform period. The recent literature stresses that firm heterogeneity gives some firms an edge over others to self-select into export markets. Apart from ownership, this paper considers firm heterogeneity and other firm-specific factors of export performance. Estimation results show that the impact of foreign ownership on export performance does not significantly differ from that of domestic firms across sectors in Indian manufacturing. Rather, firms build their international competitiveness by importing raw materials and foreign technical know-how, and by investing in research and development. Further, firm heterogeneity, measured in terms of sunk costs, significantly impacts firm-level export intensity. The study also reveals that there are ownership-specific factors that determine firm-level exports.

Convolutional Regression Framework for Human Health Prediction Under Social Influences

Understanding the propagation of human health behavior, such as smoking and obesity, and identification of the factors that control such phenomenon is an important area of research in recent years mainly because, in industrialized countries a substantial proportion of the mortality and quality of life is due to particular behavior patterns, and that these behavior patterns are modifiable. Predicting the individuals who are going to be overweight or obese in future, as overweight and obesity propagate over dynamic human interaction network, is an important problem in this area. The problem has received limited attention from the network analysis and machine learning perspective till date, though. In this work, we propose a scalable supervised prediction model based on convolutional regression framework that is particularly suitable for short time series data. We propose various schemes to model social influence for health behavior change. Further we study the contribution of the primary factors of overweight and obesity, like unhealthy diets, recent weight gains and inactivity in the prediction task. A thorough experiment shows the superiority of the proposed method over the state-of-the-art.

Does Trade Openness Increase Wage Elasticity of Labour Demand in Indian Manufacturing Industries

This chapter estimates the effect of trade openness on wage elasticity of labour demand for both production and non-production workers in aggregate as well as disaggregate manufacturing in India during the post-reform period. Econometric estimation is carried out for a panel dataset comprising fifteen disaggregated manufacturing industries for the period 1991–2010 using dynamic panel data-estimation methods. The previous studies ignored the impact of export on wage elasticity of derived labour demand, and considered import liberalisation as the sole measure of trade liberalisation. This chapter intends to include import as well as exports measures. Here, a more comprehensive measure of trade liberalisation, trade openness index, is considered for the analysis.

Generalized stacking of layerwise-trained Deep Convolutional Neural Networks for document image classification

This article presents our recent study of a lightweight Deep Convolutional Neural Network (DCNN) architecture for document image classification. Here, we concentrated on training of a committee of generalized, compact and powerful base DCNNs. A support vector machine (SVM) is used to combine the outputs of individual DCNNs. The main novelty of the present study is introduction of supervised layerwise training of DCNN architecture in document classification tasks for better initialization of weights of individual DCNNs. Each DCNN of the committee is trained for a specific part or the whole document. Also, here we used the principle of generalized stacking for combining the normalized outputs of all the members of the DCNN committee. The proposed document classification strategy has been tested on the well-known Tobacco3482 document image dataset. Results of our experimentations show that the proposed strategy involving a considerably smaller network architecture can produce comparable document classification accuracies in competition with the state-of-the-art architectures making it more suitable for use in comparatively low configuration mobile devices.

Introduction to the Special Issue on ‘Managing Balance of Payments: Fiscal and Monetary Issues’ of SAJMPF

Balance of payments across developing countries, emerging market economies in particular, behaved erratically during the past one decade. The net payments for most countries have turned into large deficits of late from a situation of ‘near-balance’ or low to moderate surplus in the recent past. In India, there was a burgeoning current account deficit in the recent past following a large trade deficit only to attain a near BOP equilibrium situation thereafter. This is evident in Figure 1. Effective domestic policy measures of import restriction and, more importantly, softening of global crude prices, led to improvements in the current account position since 2013. The overall situation can change in the future depending on changes in various parameters on which India’s current account depends, an inference which follows from an understanding of the nature of the current account deficit in India and many other similar economies. In situations of a current account deficit, the currency depreciates and brings the external accounts to a balance. Currency appreciation takes South Asian Journal of Macroeconomics and Public Finance 5(1) 2–6

FDI, Technological Choices and Spillovers in Indian Manufacturing Industries

With inflow of FDI and MNE operations in the Indian economy in the 1990s, the domestic firms had to face a very crucial issue of technology choice in the face of competition. On the one hand, technology could be imported in both embodied and disembodied form, while on the other hand, thrust could be given to develop local R&D. Again, there could also be a possibility of combining both. This chapter tries to analyse the factors influencing the firms’ technological choices across high technology, medium-high technology, medium-low technology and low-technology industries in the post reforms era. In this process, the role of the ownership of firms and technological spillovers is taken into account. A logit framework is constructed to empirically explore the technology choice determinants. Results suggest that foreign ownership and technological spillovers have significant effect on the technology choices of most Indian manufacturing industries. Dependence on imported foreign technical know-how is also evident.

Is WTO Governed Trade Regime Sufficient for Export Growth

The impact of WTO promulgated multilateral trade regime on world trade growth and welfare has remained an important issue since the formation of the institution in the mid-1990s. International trade economists have agreed to disagree on the subject. The finding of insignificant effect of WTO in promoting world trade, as empirically established by Rose (Am Econ Rev 94(1):98–114, 2004a), has been refuted by Subramanian and Wei (J Int Econ 71(1):151–175, 2007), Tomz et al. (Am Econ Rev 97(5):2005–2018, 2007), Helpman et al. (Q J Econ CXXIII(2), 2008), among others. The study investigates into whether WTO membership is sufficient for trade growth across countries. Using a gravity model framework and a dataset of 200 exporters and 234 importers, the study, controlling for trade facilitating infrastructure, time and country fixed effects along with other extended gravity variables, finds that WTO’s trade regime, even though necessary, is not sufficient for growth in trade.

Reforms, Exchange Rate Pass-Through and India's Export Prices

This paper estimates exchange rate pass-through to India’s export prices during 1960-2000. In the literature, exchange rate pass-through is mostly found to be incomplete. In this study we develop a simultaneous equation demand-supply model of export determination along the lines of an imperfect substitute model to estimate pass-through both at aggregate and disaggregate manufactured exports. This is done using the best possible econometric technique and a time comparable dataset. The study shows high, but incomplete, exchange rate pass-through into India’s aggregate export prices. The degree of exchange rate pass-through is found to vary across product groups.