Cochin University of Science and Technology

About: Cochin University of Science and Technology is a education organization based out in Kochi, Kerala, India. It is known for research contribution in the topics: Thin film & Natural rubber. The organization has 5382 authors who have published 7690 publications receiving 103827 citations. The organization is also known as: CUSAT & Cochin University.

An investigation into observational characteristics of rainfall and temperature in Central Northeast India—a historical perspective 1889–2008

Abstract: Mann–Kendall non-parametric test was employed for observational trend detection of monthly, seasonal and annual precipitation of five meteorological subdivisions of Central Northeast India (CNE India) for different 30-year normal periods (NP) viz. 1889–1918 (NP1), 1919–1948 (NP2), 1949–1978 (NP3) and 1979–2008 (NP4). The trends of maximum and minimum temperatures were also investigated. The slopes of the trend lines were determined using the method of least square linear fitting. An application of Morelet wavelet analysis was done with monthly rainfall during June–September, total rainfall during monsoon season and annual rainfall to know the periodicity and to test the significance of periodicity using the power spectrum method. The inferences figure out from the analyses will be helpful to the policy managers, planners and agricultural scientists to work out irrigation and water management options under various possible climatic eventualities for the region. The long-term (1889–2008) mean annual rainfall of CNE India is 1,195.1 mm with a standard deviation of 134.1 mm and coefficient of variation of 11%. There is a significant decreasing trend of 4.6 mm/year for Jharkhand and 3.2 mm/day for CNE India. Since rice crop is the important kharif crop (May–October) in this region, the decreasing trend of rainfall during the month of July may delay/affect the transplanting/vegetative phase of the crop, and assured irrigation is very much needed to tackle the drought situation. During the month of December, all the meteorological subdivisions except Jharkhand show a significant decreasing trend of rainfall during recent normal period NP4. The decrease of rainfall during December may hamper sowing of wheat, which is the important rabi crop (November–March) in most parts of this region. Maximum temperature shows significant rising trend of 0.008°C/year (at 0.01 level) during monsoon season and 0.014°C/year (at 0.01 level) during post-monsoon season during the period 1914–2003. The annual maximum temperature also shows significant increasing trend of 0.008°C/year (at 0.01 level) during the same period. Minimum temperature shows significant rising trend of 0.012°C/year (at 0.01 level) during post-monsoon season and significant falling trend of 0.002°C/year (at 0.05 level) during monsoon season. A significant 4–8 years peak periodicity band has been noticed during September over Western UP, and 30–34 years periodicity has been observed during July over Bihar subdivision. However, as far as CNE India is concerned, no significant periodicity has been noticed in any of the time series.

Lithospheric structure below the eastern Arabian Sea and adjoining West Coast of India based on integrated analysis of gravity and seismic data

Abstract: Four uniformly spaced regional gravity traverses and the available seismic data across the western continental margin of India, starting from the western Indian shield extending into the deep oceanic areas of the eastern Arabian Sea, have been utilized to delineate the lithospheric structure. The seismically constrained gravity models along these four traverses suggest that the crustal structure below the northern part of the margin within the Deccan Volcanic Province (DVP) is significantly different from the margin outside the DVP. The lithosphere thickness, in general, varies from 110–120 km in the central and southern part of the margin to as much as 85–90 km below the Deccan Plateau and Cambay rift basin in the north. The Eastern basin is characterised by thinned rift stage continental crust which extends as far as Laxmi basin in the north and the Laccadive ridge in the south. At the ocean–continent transition (OCT), crustal density differences between the Laxmi ridge and the Laxmi basin are not sufficient to distinguish continental as against an oceanic crust through gravity modeling. However, 5-6 km thick oceanic crust below the Laxmi basin is a consistent gravity option. Significantly, the models indicate the presence of a high density layer of 3.0 g/cm3 in the lower crust in almost whole of the northern part of the region between the Laxmi ridge and the pericontinental northwest shield region in the DVP, and also below Laccadive ridge in the southern part. The Laxmi ridge is underlain by continental crust upto a depth of 11 km and a thick high density material (3.0 g/cm3) between 11–26 km. The Pratap ridge is indicated as a shallow basement high in the upper part of the crust formed during rifting. The 15 –17 km thick oceanic crust below Laccadive ridge is seen further thickened by high density underplated material down to Moho depths of 24–25 km which indicate formation of the ridge along Reunion hotspot trace.

High-Performance CNN Accelerator on FPGA Using Unified Winograd-GEMM Architecture

Abstract: Deep neural networks have revolutionized a variety of applications in varying domains like autonomous vehicles, weather forecasting, cancer detection, surveillance, traffic management, and so on. The convolutional neural network (CNN) is the state-of-the-art technique for many machine learning tasks in the image and video processing domains. Deployment of CNNs on embedded systems with lower processing power and smaller power budget is a challenging task. Recent studies have shown the effectiveness of field-programmable gate array (FPGA) as a hardware accelerator for the CNNs that can deliver high performance at low power budgets. Majority of computations in CNNs involve 2-D convolution. Winograd minimal filtering-based algorithm is the most efficient technique for calculating convolution for smaller filter sizes. CNNs also consist of fully connected layers that are computed using general element-wise matrix multiplication (GEMM). In this article, we propose a unified architecture named UniWiG, where both Winograd-based convolution and GEMM can be accelerated using the same set of processing elements. This approach leads to efficient utilization of FPGA hardware resources while computing all layers in the CNN. The proposed architecture shows performance improvement in the range of $1.4\times $ to $4.02\times $ with only 13% additional FPGA resources with respect to the baseline GEMM-based architecture. We have mapped popular CNN models like AlexNet and VGG-16 onto the proposed accelerator and the measured performance compares favorably with other state-of-the-art implementations. We have also analyzed the vulnerability of the accelerator to the side-channel attacks. Preliminary investigations show that the UniWiG architecture is more robust to memory side-channel attacks than direct convolution-based techniques.