AR_final file_2018-19

a flat universe. It is shown that a universe with non-linear EoS is effectively a cosmological model with a composition of three different fluids. In the original model, one of the drawbacks is that the components of the fluids are fixed for a definite EoS parameter. Subsequently , considering interactions among the fluids it is found that a viable cosmo- logical scenario can be obtained. In the presence of interactions among the fluids, which are permitted by the EoS, it is found to accommodate realistic cosmological models in accordance with observa- tions. It has also been shown that the initial static Einstein phase required in the EU model is permit- ted with a dynamical wormhole in the framework of massive gravity This study has been done in col- laboration with A. S. Majumdar. Relativistic star in higher dimensions with Finch - Skea geometry Relativistic solutions of compact objects in hydro- dynamical equilibrium with Finch - Skea (FS) met- ric are obtained in higher dimensions. They are studied to obtain seller models for compact stars in the usual four and in higher dimensions. We study variation of different physical parameters in- side the star. Considering the known stars, we ob- tain stellar models. It is noted that a compact star in 4 - dimensions with Finch - Skea geometry al- ways admits a star with isotropic pressure, however in higher dimensions it always admits anisotropic star. The plausibility of such stars are, also stud- ied here. This study has been done in collaboration with Sagar Dey. Ninan Sajeeth Philip Machine learning applications in the time domain astronomy The Catalina Real Time Transient (CRTS) survey is followed by the ZTF survey, which itself is to be the forerunner of some of the major time do- main astronomy surveys that are planned for the coming decade. We study on using deep learning tools to automatically detect and follow up tran- sient events in these surveys. Currently, we work on a method developed by Ashish Mahabal et al., at Caltech, where they converted light curves ob- tained from CRTS northern periodic variables to images using a two dimensional mapping called dm- dt maps. This allows us to work with images that are more stable than the dynamic fluctuations usu- ally observed in their lightcurves. It was noted that instead of training the classifier on the entire data of 17 classes, if we do a binary classification of one against all the rest in rotation, the accuracy in pre- diction becomes more reliable. The reason for this is the imbalance in the class distributions that bi- ases the classifier, that just try to minimise the error, towards the objects that have higher repre- sentation in the dataset. It was observed that the binary classification could improve the prediction accuracy of 53% in 17 classes to the order of 97% and above when addressed as a binary problem of one against all in rotation. This study has been done in collaboration with G. Sindhu, and Linn Abraham. Application of machine learning in biosciences (ABCD portal) The Astronomy Biology Combined Data ABCD portal is a joint research programme, funded by NKN for porting some of the tools developed by astronomers to benefit the bioscience community. Species diversity is one of the key parameters biol- ogists use to understand the health of any ecosys- tem. We could develop a prototype machine learn- ing tool for the automated detection and identifica- tion of different freshwater fish from video footings taken with underwater surveillance cameras. The tool could also select and track individual fish with appropriate tracking for their behavioural studies. When working with underwater surveillance video footings, background objects create a lot of un- certainties in the automated classification of the species. We have developed a tool to do auto- mated background subtraction from video footings by considering frame differences much like what as- tronomers do for sky subtraction. This study has been done in collaboration with Geetha Paul, Bles- son George, and Linn Abraham. Anirudh Pradhan Diagnosing Tsallis holographic dark energy models with statefinder and ω − ω ′ pair A useful method, known as statefinder diagnos- tic, which may differentiate one dark energy model from others is applied in this work to a holo- graphic dark energy model from Tsallis entropy, called the Tsallis holographic dark energy (THDE) model. The evolutionary trajectories are plotted ( 207 )