You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$1,800 left to reach the funding goal.
During my PhD at IISc Bengaluru, our team developed Adaptive Multiscale Operator Correction (AMOC), a new framework for scientific machine learning.
Neural operators are fast at solving PDEs, but they often miss sharp peaks, boundary layers, and other high-frequency details. AMOC improves them by adding physics-constrained corrections in an adaptive Fourier–wavelet subspace, targeting exactly where standard models struggle.
On heat, Poisson, Darcy, and Helmholtz problems, AMOC reduced errors by up to four orders of magnitude compared with standard baselines, especially for out-of-distribution cases. Our paper has been accepted at NeurIPS 2026, and the code is publicly available.
My immediate goal is to present our accepted paper in person at NeurIPS 2026 in Sydney and discuss it with researchers in scientific machine learning, computational mathematics, and physics-informed AI.
The broader goal is to build fast AI-based PDE solvers that do not lose important physical details. Feedback from NeurIPS will help us improve AMOC, especially its computational cost, choice of spectral bases, and memory use for 3D wavelet representations.
After the conference, I plan to use these ideas at IISc to develop better hybrid Fourier–wavelet models and explore residual-aware basis selection for applications such as wave propagation and atmospheric dynamics.
I am requesting support to travel from Bengaluru to Sydney and present our accepted NeurIPS 2026 paper.
I have received NeurIPS Financial Assistance, which covers my conference registration and accommodation. Therefore, I am seeking support mainly for travel and remaining local expenses.
Round-trip airfare: $1,200
Meals and incidentals: $450
Local transportation and poster printing: $150
Total requested: $1,800
I am also exploring institutional travel support and other funding options. However, no additional travel funding for this trip has been confirmed yet. Any support received would directly help me attend NeurIPS in person and present our work to the research community.
Our team includes me, Subham Patel; Himanshu Pandey; and our mentor and corresponding author, Dr. Ratikanta Behera, from the Departments of Computational and Data Sciences and Mathematics at IISc.
As first author, I led the formulation of the Tikhonov-regularized wavelet projection, developed the AMOC architecture in PyTorch, ran multi-GPU physics-informed experiments, and derived closed-form derivatives using Hermite polynomials.
My work focuses on neural operators, physics-informed methods, and computational mathematics for parametric PDEs. Our accepted NeurIPS paper and open-source code show that we can build, evaluate, and share reproducible research.
The research will continue even if I cannot travel: the paper is accepted and the code is public. However, without support, I may miss the chance to present the work in person, receive technical feedback, and build collaborations.
Attending NeurIPS would also help me represent IISc, strengthen my research training, and bring useful ideas back to our ongoing work on 3D wavelet methods and adaptive spectral corrections.
My PhD is supported by a fellowship through the Axis Bank Centre for Mathematics and Computing at IISc and the Ministry of Education, Government of India. This covers my regular stipend but not international conference travel.