Indian Institute of Technology Gandhinagar

We study the query complexity of finding tournament solutions in near-transitive tournaments, which are tournaments obtained by flipping a small number of edges in a transitive tournament. While general tournaments require Ω(n²) queries for many solution concepts such as Copeland, Top Cycle, and Uncovered Set, we show that we can do better for near-transitive tournaments in several scenarios. In particular, we introduce three query models: the standard model, where queries return orientations in the input tournament; the partial-flip model, where queries return orientations from the underlying transitive tournament and we have access to edge-flip queries; and the full-flip model, which additionally provides vertex-flip queries that indicate whether a vertex is incident to a flipped edge.

India’s agriculture sector is vital for employment and the economy, but it faces several challenges despite digital initiatives. The sector is unorganised, fragmented, and lacks integrated social networking, hindering collaboration. This study used a qualitative, exploratory design with semi-structured interviews of ten agri stakeholders. The thematic analysis identified four key barriers: unreliable information channels, weakening institutional trust, unstable income generation, and market inefficiencies that disempower farmers. This research contributes to a deeper understanding of the digital needs of Indian farmers and their current agri ecosystem. This study further explores digital interventions that can be integrated into the current agri-ecosystem. Future researchers can build upon the findings of this work to develop prototype ideas and test them in the market, thereby gaining an in-depth understanding of the impact of such interventions.

Optimal control of hybrid vehicles involving a battery and a combustion engine is necessary for efficient energy management, and this optimal control problem (OCP) can be formulated as a mixed-integer program (MIP). The real-time deployment of MIPs is challenging due to their computational complexity. As the number of integer variables increases, the solution space increases exponentially, requiring computationally expensive methods to solve. Although machine learning-based strategies help reduce computation time, they provide approximate, but not exact, solutions to MIPs and may even involve expensive one-time training, employing heavy computational resources. In this work, a multiparametric (mp) programming framework is used to improve the computation time of an OCP for hybrid vehicles while getting the exact solution; furthermore, the mp-programming framework allows for implementation through low-cost hardware like a chip without needing a computer. In the mp-programming approach, the solution is expressed as a set of piecewise linear functions in terms of the uncertain parameters that change with time. During online calculations, the optimal solution is determined through a point location search procedure. This framework is compared with the state-of-the-art branch-and-bound method in the Gurobi solver and a neural network solver model developed in this work.

Brain-computer interface (BCI) systems facilitate unique communication between humans and computers, benefiting severely disabled individuals. Despite decades of research, BCIs are not fully integrated into clinical and commercial settings. It's crucial to assess and explain BCI performance, offering clear explanations for potential users to avoid frustration when it doesn't work as expected. This work investigates the efficacy of different deep learning and Riemannian geometry-based classification models in the context of motor imagery (MI) based BCI using electroencephalography (EEG). We then propose an optimal transport theory-based approach using earth mover's distance (EMD) to quantify the comparison of the feature relevance map with the domain knowledge of neuroscience. For this, we utilized explainable AI (XAI) techniques for generating feature relevance in the spatial domain to identify important channels for model outcomes. Three state-of-the-art models are implemented - 1) Riemannian geometry-based classifier, 2) EEGNet, and 3) EEG Conformer, and the observed trend in the model's accuracy across different architectures on the dataset correlates with the proposed feature relevance metrics. The models with diverse architectures perform significantly better when trained on channels relevant to motor imagery than data-driven channel selection. This work focuses attention on the necessity for interpretability and incorporating metrics beyond accuracy, underscores the value of combining domain knowledge and quantifying model interpretations with data-driven approaches in creating reliable and robust Brain-Computer Interfaces (BCIs).

The ĐatalLJtiĐ perforŵaŶĐe of Pd ŶaŶopartiĐles supported oŶ ϒ-Al2O3 aŶd α-Al2O3 have ďeeŶ iŶǀestigated iŶ the seleĐtiǀe hLJdrogeŶatioŶ of aĐetLJleŶe. ϒ-Al2O3 aŶd α-Al2O3 were synthesized by solution combustion using Al(NO3)3.9H2O precursor as well as sol gel using Al(OC4H9)3 and AlCl3.6H2O precursors. Pd dispersion on the Al2O3 support (1wt% Pd/ Al2O3) was accomplished via chemical reduction method using formaldehyde as reducing agent at 80°C. Low temperature synthesis of Pd/ Al2O3 catalyst prevented the doping of the compound and ensured the presence of Pd inmetallic state. Effect of temperature and flow rate on the catalytic activity of the synthesized Pd/ Al2O3 catalysts have been explored in the present work. It was observed that ethylene selectivity increases as surface area decreases in the Pd/ Al2O3 catalysts synthesized by solution combustion and sol gel (using Al(OC4H9)3 precursor) technique. Lower is the surface area, lower is the Pd dispersion on the support and lower are the active sites available for direct ethane formation. However, this is not universally true as can be noticed in the case of Pd/ Al2O3 catalysts synthesized by sol gel technique using AlCl3.6H2O precursor. In this case, irrespective of the surface area ethylene yield remained zero. Preparation method of support modified the surface nature of the catalyst which played another crucial role in ethylene selectivity. CO2 temperature programmed desorption (TPD) studies suggested that lower basicity of the Al2O3 surface resulted in better ethylene selectivity for both solution combustion and sol gel method. Moreover, the crystallite size of α-Al2O3 particles is not an important parameter affecting ethylene selectivity. Pd/ α-Al2O3 catalyst synthesized by solution combustion method using Al(NO3)3.9H2O precursor at 1200°C shows the best ethylene selectivity during the entire temperature range from RT to 125°C. 100% conversion as well as good selectivity is simultaneously obtained at 125°C which is a superior performance as compared to most of the existing catalysts in literature. Pd/ ϒ-Al2O3 catalyst synthesized by solution combustion method using Al(NO3)3.9H2O precursor at 700°C is a moderately good catalyst in terms of ethylene selectivity. It shows a decrease in conversion with the increase in total flow rate of reactant gases. However, flow rate seems to have a negligible impact on ethylene selectivity as well as yield. We may extend this result to other catalysts as well.