Research

My research develops tensor and multilinear methods as a unifying tool for making quantum and hybrid quantum-classical machine learning practical under NISQ-era constraints. I focus on methods that reduce circuit evaluations, parameter overhead, memory footprint, and deployment friction while staying experimentally grounded.

Current Focus

Efficient QML Training and Deployment

I design variational training and readout schemes that are deployable on real noisy hardware. One example is VQBR, a variational quantum Bayesian regression framework that uses measurement-based MAP direction recovery to reduce circuit evaluation and readout overhead.

Tensor-Structured Classical and Hybrid Architectures

I build low-rank and Tucker-factored multilinear transformation layers for compact neural networks. This includes low-rank multilinear transformations and quantum tensor contraction layers that reduce parameter counts without sacrificing accuracy.

Quantum Algorithms for Classical Problems

I study quantum primitives such as HHL-based linear solvers and quantum tensor-network formulations for classical computations, including tensor decomposition and linear regression.

Improving Existing QML Frameworks

I re-architect components of established QML frameworks with tensor structure. For example, in quantum reservoir computing, I study Tucker-factored multilinear readouts as a parameter-efficient alternative to flat linear readouts.

Experimental Practice

I validate methods through reproducible simulator experiments, GPU workflows, and accessible quantum hardware when appropriate. My main tools include Qiskit, PennyLane, Cirq, CUDA-Q, PyTorch, TensorFlow, and scikit-learn.

Collaboration

I am open to research conversations around quantum machine learning, hybrid quantum-classical learning, tensor methods, compact neural networks, and quantum algorithms for classical computation. You can reach me at tien.nguyen@utsa.edu.