A5101683208- Complexity and Algorithms in Graphs
- Advanced Graph Theory Research
- Machine Learning and Data Classification
- Optimization and Search Problems
- Immunotherapy and Immune Responses
- Explainable Artificial Intelligence (XAI)
- Machine Learning in Materials Science
- Limits and Structures in Graph Theory
- Computability, Logic, AI Algorithms
- Coding theory and cryptography
- RNA Interference and Gene Delivery
- Logic, programming, and type systems
- Auction Theory and Applications
- Quantum Computing Algorithms and Architecture
- Vehicle Routing Optimization Methods
- T-cell and B-cell Immunology
- Microgrid Control and Optimization
- Mathematical Biology Tumor Growth
- Cryptography and Data Security
- Bacterial Infections and Vaccines
- Integrated Circuits and Semiconductor Failure Analysis
- Engineering and Test Systems
- Urban and Freight Transport Logistics
- Advanced Optical Network Technologies
- Viral Infectious Diseases and Gene Expression in Insects
The University of Texas at Austin
2023-2024
Indraprastha Institute of Information Technology Delhi
2019
Massachusetts Institute of Technology
2002-2019
Vellore Institute of Technology University
2017
Novartis (United States)
2011-2014
Brown University
2010-2011
Indian Institute of Technology Indore
2002
The inaccuracy of neural network models on inputs that do not stem from the distribution underlying training data is problematic and at times unrecognized. Uncertainty estimates model predictions are often based variation in produced by a diverse ensemble applied to same input. Here we describe Maximize Overall Diversity (MOD), an approach improve ensemble-based uncertainty encouraging larger overall diversity across all possible inputs. We apply MOD regression tasks including 38 Protein-DNA...
Test set contamination, wherein test data from a benchmark ends up in newer model's training set, is well-documented obstacle for fair LLM evaluation and can quickly render benchmarks obsolete. To mitigate this, many recent crowdsource new prompts evaluations human or judges; however, these introduce significant biases, break down when scoring hard questions. In this work, we LLMs designed to be immune both contamination the pitfalls of judging crowdsourcing. We release LiveBench, first that...
Materials that effectively deliver protein antigens together with activating ligands to antigen-presenting cells are sought for improved nonviral vaccines. To this end, we synthesized protein-loaded poly(ethylene glycol) (PEG)-based hydrogel particles by cross-linking PEG within the polymer-rich phase of an emulsion formed a oxide-b-propylene oxide-b-ethylene oxide) triblock copolymer in saturated aqueous salt solution. These (500-nm diameter) contained high levels encapsulated (∼75% dry...
Testing plays a pivotal role in ensuring software quality, yet conventional Search Based Software (SBST) methods often struggle with complex units, achieving suboptimal test coverage. Recent work using large language models (LLMs) for generation have focused on improving quality through optimizing the context and correcting errors model outputs, but use fixed prompting strategies that prompt to generate tests without additional guidance. As result LLM-generated suites still suffer from low...
It is well-known that Resolution proofs can be efficiently simulated by Sherali–Adams (SA) proofs. We show, however, any such simulation needs to exploit huge coefficients: cannot SA when the coefficients are written in unary. also show Reversible (a variant of MaxSAT Resolution) Nullstellensatz (NS). These results have consequences for total NP search problems. First, we characterise classes PPADS, PPAD, SOPL unary-SA, unary-NS, and Resolution, respectively. Second, that, relative an...
Subunit vaccines are typically poorly immunogenic when administered alone, and require adjuvants for robust responses. Triggering TLRs to boost antigen-specific adaptive immunity is an attractive approach increase the potency quality of vaccines. However, recent reports suggest that alterations in TLR expression associated with pathogenesis inflammatory autoimmune diseases. To compare genetic studies adjuvant studies, we examined whether stimulation through a agonist induces or increases...
It is well-known that Resolution proofs can be efficiently simulated by Sherali-Adams (SA) proofs. We show <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> , however, any such simulation needs to exploit huge coefficients: cannot SA when the coefficients are written in unary. also Reversible (a variant of MaxSAT Resolution) Nullstellensatz (NS). These results have consequences for total NP search problems. First, we characterise classes...
Local explanation frameworks aim to rationalize particular decisions made by a black-box prediction model. Existing techniques are often restricted specific type of predictor or based on input saliency, which may be undesirably sensitive factors unrelated the model's decision making process. We instead propose sufficient subsets that identify minimal features whose observed values alone suffice for same reached, even if all other feature missing. General principles globally govern...
We exhibit an unambiguous -DNF (disjunctive normal form) formula that requires conjunctive form width , which is optimal up to logarithmic factors. As a consequence, we get near-optimal solution the Alon–Saks–Seymour problem in graph theory (posed 1991), asks, How large gap can there be between chromatic number of and its biclique partition number? Our result also known imply several other improved separations query/communication complexity, learning theory, automata theory.
A kinetic model has been developed to study cancer growth. Cancer growth considered as interaction between various independent but interacting compartments. The considers cell and metastasis resulting in the formation of new tumor masses. Using certain representative parameter values, modeled absence presence therapies. Based on this analysis, critical parameters involved development have identified. This may thus be useful studying designing a therapy using data obtained from specific vitro...
This paper presents a model of stand-alone solar energy conversion system based on synchronous machine working as condenser in overexcited state. The proposed consists Synchronous Condenser, DC/DC boost converter whose output is fed to the field SC. supplied by modelled panel and day time variable irradiance during simulation time. also has one alternate source rechargeable batteries for when falls below threshold value. Also excess power produced there ample divided two parts while other...
The generalized pigeonhole principle says that if tN + 1 pigeons are put into N holes then there must be a hole containing at least t pigeons. Let t-PPP denote the class of all total NP-search problems reducible to finding such t-collision We introduce new hierarchy classes defined by t-PPP. In addition being natural in TFNP, we show and above related notion multi-collision resistance cryptography, contain problem underlying breakthrough average-case quantum advantage result shown Yamakawa &...
.We show \(\mathsf{EOPL}=\mathsf{PLS}\cap \mathsf{PPAD}\). Here the class \(\mathsf{EOPL}\) consists of all total search problems that reduce to End-of-Potential-Line problem, which was introduced in works by Hubáček and Yogev (SICOMP 2020) Fearnley et al. (JCSS 2020). In particular, our result yields a new simpler proof breakthrough collapse \(\mathsf{CLS}=\mathsf{PLS}\cap \mathsf{PPAD}\) (STOC 2021). We also prove companion \(\mathsf{SOPL}=\mathsf{PLS}\cap \mathsf{PPADS}\), where...
We exhibit an unambiguous <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$k$</tex> -DNF formula that requires CNF width xmlns:xlink="http://www.w3.org/1999/xlink">$\tilde\Omega(k^{2})$</tex> , which is optimal up to logarithmic factors. As a consequence, we get near-optimal solution the Alon–Saks–Seymour problem in graph theory (posed 1991), asks: How large gap can there be between chromatic number of and its biclique partition number? Our...
It is well-known that Resolution proofs can be efficiently simulated by Sherali-Adams (SA) proofs. We show, however, any such simulation needs to exploit huge coefficients: cannot SA when the coefficients are written in unary. also show Reversible (a variant of MaxSAT Resolution) Nullstellensatz (NS). These results have consequences for total NP search problems. First, we characterise classes PPADS, PPAD, SOPL unary-SA, unary-NS, and Resolution, respectively. Second, that, relative an...
The inaccuracy of neural network models on inputs that do not stem from the training data distribution is both problematic and at times unrecognized. Model uncertainty estimation can address this issue, where estimates are often based variation in predictions produced by a diverse ensemble applied to same input. Here we describe Maximize Overall Diversity (MOD), straightforward approach improve ensemble-based encouraging larger overall diversity across all possible might be encountered...