Dhruv Patel

Dhruv Patel

Postdoctoral Fellow

Stanford University

About me

I am currently the Stephen Timoshenko Distinguished Postdoctoral Fellow in the Mechanics and Computation Group at Stanford University. Earlier I completed my Ph.D. in the Aerospace and Mechanical Engineering department at USC under the supervision of Prof. Assad Oberai.

My research interest lies at the intersection of physics-based and data-driven modeling, deep learning, inverse problems, and uncertainty quantification with applications to computational science and engineering, computer vision, and medical imaging. I am particularly interested in the question of how to optimally combine physics-guided models with modern machine learning algorithms to develop more data efficient and accurate hybrid models.

Interests

  • Scientific Machine Learning
  • Bayesian Inferernce
  • Deep Learning
  • Generative Modeling
  • Uncertainty Quantification

Education

  • PhD in Mechanical Engineering

    University of Southern California

  • MTech in Applied Mechanics

    Indian Institute of Technology, Delhi

  • BE in Mechanical Engineering

    L. D. College of Engineering, Ahmedabad

News

Feb 2022: New preprint on Efficacy and Generalizability of Conditional GANs for Solving Physics-based Bayesian Inverse Problems

Dec 2021: Will present a poster at Deep inverse workshop at NeurIPS on Conditional GAN-based Bayesian inversion.

Dec 2021: Giving a talk at AGU21 on efficient subsurface inversion using deep learning and multi-fidelity modeling.

Nov 2021: New preprint on Multi-fidelity Hamiltonian Monte Carlo method - to be presented at AAAI'21 (symposium on science guided AI).

Oct 2021: Giving an invited talk at the ERE gradute seminar, School of Earth, Energy, and Environmental Sciences, Stanford.

Recent Publications

Solution of Physics-based Bayesian Inverse Problems with Deep Generative Priors

Inverse problems are notoriously difficult to solve because they can have no solutions, multiple solutions, or have solutions that vary …

Machine Learning Based Predictors for COVID-19 Disease Severity

Predictors of the need for intensive care and mechanical ventilation can help healthcare systems in planning for surge capacity for …

Bayesian Inference in Physics-driven Problems with Adversarial Priors

Generative adversarial networks (GANs) have found multiple applications in the solution of inverse problems in science and engineering. …

Probabilistic Recovery of Missing Phase Images in Contrast-Enhanced CT

Contrast-Enhanced CT (CECT) imaging is used in the diagnosis of renal cancer and planning of surgery. Often, some CECT phase images are …

Deep Learning-based Detection, Classification, and Localization of Defects in Semiconductor Processes

Defects in semiconductor processes can limit yield, increase overall production cost, and also lead to time-dependent critical …

GAN-based Priors for Quantifying Uncertainty

Bayesian inference is used extensively to quantify the uncertainty in an inferred field given the measurement of a related field when …

Recent Posts

How deep learning combined with bio-mechanics can lead to accurate and non-invasive diagnosis of breast cancer?

Summary: Breast ultrasound elastography is a non-invasive technique used by radiologists to help detect and diagnose cancer and other diseases by evaluating a lesion’s stiffness in a non-invasive way. Researchers identified the critical role machine learning can play in making this technique more efficient and accurate in diagnosis.

Contact

  • Building 520, 440 Escondido Mall, Stanford, CA 94305