Browsing by Author "Nair, R.R."

Now showing 1 - 3 of 3

Co-digestion of mushroom compost with switchgrass using solid-state anaerobic digester
(ICE Publishing, 2023) Nair, R.R.; Thalla, A.K.; Nair, V.V.
Spent mushroom compost (SMC) already broken down into smaller particles by fungal action is an ideal material for producing biogas. Two cycles of five solid-state anaerobic digesters (SS-ADs) with different mix-ratio of SMC and switchgrass (SG) were operated at feedstock-to-effluent ratio of 2 at a temperature 35 ± 2°C. The total solids concentration of the digester was kept at 17%. Initial biogas production observed during the start-up of the digester confirmed the presence of readily available extractives for digestion. In the first cycle, the highest methane yield was observed in SMC 0 (0% SMC + 100% SG) of 28.82 l/kg VS/d and the lowest yield was observed in SMC 4 (100% SMC + 0% SG) as 10.32 l/kg VS/d. The substrate containing 100% SG (SMC 0) recorded the highest cumulative biogas yield of 295.43 l/kg VS in 63 days. The digesters with higher SMC fraction showed lower methane production, low pH value and high volatile fatty acids content upon decomposition. The SS-ADs having SMC/SG of 50 : 50 showed more than 2 times methane production in comparison with SS-ADs having SMC as sole substrate. An estimation of volumetric productivity also established a linear relationship with the SMC/SG ratio. © 2023 Emerald Publishing Limited: All rights reserved.
Hidden Markov Model for Hard Disk Drive Failure Detection
(Institute of Electrical and Electronics Engineers Inc., 2024) Harish, A.; Prakash, G.; Nair, R.R.; Iyer, V.B.; Mohan, B.R.; Das, M.
Understanding disk failures is crucial for both disk manufacturers and users, enabling the production of more dependable disk drives and the establishment of robust storage systems. Detecting disk failure has been found to be facilitated by the use of observable disk properties, especially those provided by the Self-Monitoring and Reporting Technology (SMART) system. In our paper, we leverage the capabilities of the SMART time series dataset to achieve an overall accuracy of 92% in disk failure detection. Â© 2024 IEEE.
Refining LLMs with Reinforcement Learning for Human-Like Text Generation
(Institute of Electrical and Electronics Engineers Inc., 2024) Harish, A.; Prakash, G.; Nair, R.R.; Iyer, V.B.; Anand Kumar, M.
Large Language Models (LLMs) are used widely for tasks involving text generation such as dialogue summarization and creative writing. The generated text often appears unnatural, and this text can easily be distinguished from natural language. In this paper, we leverage the capabilities of Reinforcement Learning to fine-tune LLMs so as to produce text that resembles human language. We have applied the Proximal Policy Optimization algorithm to fine tune a FLAN-T5 LLM for a dialogue summarization task. Â© 2024 IEEE.