Research

Four streams, one community

Our research spans the theoretical foundations of machine learning, the emerging discipline of quantum machine learning, and the practical engineering of hybrid quantum-classical systems — held together by a weekly theory reading group that keeps the entire club grounded in first principles.

Research Streams

Active research areas

Each stream meets regularly, runs collaborative projects, and produces technical output for the community.

Machine Learning Theory

Our ML Theory stream provides the mathematical foundations that underpin both classical and quantum learning. We study the statistical limits of learning algorithms, the geometry of loss landscapes, and the representational capacity of modern model families. This stream serves as the theoretical backbone of the club, ensuring that our quantum-focused work is grounded in rigorous learning-theoretic principles.

Key Topics

  • PAC learning and VC dimension
  • Rademacher and Gaussian complexity
  • Generalization bounds for neural networks
  • Implicit regularization and overparameterization
  • Optimization landscapes and saddle-point methods
  • Information-theoretic lower bounds
  • Kernel methods and the reproducing kernel Hilbert space

Open Questions

  • How do double-descent phenomena in overparameterized models connect to quantum circuit depth?
  • Can PAC-learning bounds be tightened for structured hypothesis classes arising in physics?
  • What role does algorithmic stability play in generalization for stochastic gradient methods on non-convex objectives?

Quantum Machine Learning

Quantum Machine Learning (QML) investigates how quantum mechanical phenomena — superposition, entanglement, and interference — can be leveraged to learn from data more efficiently than classical algorithms permit. Our QML stream covers both provable speedup results and near-term heuristic approaches, maintaining a critical perspective on which claims of quantum advantage are rigorously supported.

Key Topics

  • Variational quantum eigensolvers (VQE) and QAOA
  • Parameterized quantum circuits as function approximators
  • Quantum kernel methods and feature maps
  • Quantum PCA and singular value estimation
  • Dequantization and the limits of quantum speedups
  • Quantum generalization theory and sample complexity
  • Noise resilience and error mitigation in NISQ models

Open Questions

  • Which learning tasks have provable quantum speedups that survive dequantization?
  • How does barren plateau prevalence scale with circuit architecture and problem structure?
  • Can quantum advantage be demonstrated on real-world (non-synthetic) datasets?

Hybrid Classical-Quantum Systems

In the NISQ era, neither fully classical nor fully quantum approaches are likely to be optimal across all tasks. Our Hybrid Systems stream explores architectures and algorithms that delegate computation between classical and quantum processors in a principled way, exploiting the comparative advantages of each. We are particularly interested in scientific computing applications where physical structure informs model design.

Key Topics

  • Quantum-classical variational algorithms
  • Quantum transfer learning and fine-tuning
  • Quantum feature extraction with classical classifiers
  • Tensor network methods bridging classical and quantum
  • Physics-informed neural networks with quantum layers
  • Hybrid optimization for combinatorial problems
  • Co-design of quantum circuits and classical post-processing

Open Questions

  • How should the boundary between quantum and classical computation be chosen to maximize practical advantage?
  • Can hybrid architectures leverage quantum noise as a computational resource rather than treating it purely as an error source?
  • What are the communication complexity costs of hybrid protocols, and how do they affect end-to-end speedups?

Theory Reading Group

The Theory Reading Group meets weekly to work through foundational and frontier papers in quantum information, computational complexity, and statistical learning. Attendance is open to all MIC members. Sessions are structured as collaborative discussions rather than lectures, with a designated lead summarizing the paper and guiding questions. The group has worked through textbooks by Nielsen & Chuang, Shalev-Shwartz & Ben-David, and Schuld & Petruccione.

Key Topics

  • Quantum information theory and quantum channel capacity
  • Computational complexity of quantum simulation
  • Statistical mechanics of learning systems
  • Random matrix theory and high-dimensional probability
  • Quantum Shannon theory and quantum error correction
  • Connections between statistical physics and deep learning

Open Questions

  • How does the No-Free-Lunch theorem generalize to the quantum setting?
  • What complexity-theoretic separations exist between quantum and classical learners?
  • Can techniques from random matrix theory better characterize quantum circuit expressivity?

How It Works

From reading group to research output

MIC research follows a deliberate progression designed to take members from paper consumer to paper producer.

1

Foundational reading

New members start with the Theory Reading Group, working through textbook-level material and seminal papers to build shared vocabulary.

2

Stream membership

Once fluent in fundamentals, members join a focused research stream. They read frontier papers, attend weekly meetings, and contribute to discussions.

3

Project contribution

Members are matched with active projects. They contribute code, analysis, literature review, or experiments — with mentorship from research leads.

4

Research output

Mature contributors produce technical reports, workshop submissions, or open-source tools. Research leads guide writing, reviewing, and submission.

Projects

Current and completed projects

Open-source tools, benchmarks, and research experiments maintained by MIC members.

GitHub org
ActiveSince Jan 2024

QML Benchmarking Suite

An open-source Python library for benchmarking quantum machine learning algorithms against classical baselines. The suite provides standardized datasets, evaluation protocols, and visualization tools for researchers studying near-term quantum advantage. Compatible with PennyLane, Qiskit, and Cirq backends.

PythonPennyLaneQiskitBenchmarkingOpen Source

Leads: Priya Nair, Raj Patel

Contributors: Daniel Kim, Yuki Tanaka, Leon Fischer

ActiveSince Jun 2024

Variational Autoencoder on Quantum Hardware

An exploration of quantum variational autoencoders (QVAEs) implemented on IBM Quantum hardware. The project investigates whether quantum latent spaces provide meaningful representational advantages for structured datasets, with a focus on molecular fingerprint compression for drug discovery applications.

QMLQiskitIBM QuantumAutoencodersDrug Discovery

Leads: Sofia Martinez, Tobias Richter

Contributors: Amara Diallo, Nadia Hassan

ActiveSince Sep 2023

Open QML Curriculum

A freely available, structured learning curriculum for quantum machine learning, developed iteratively through three cohorts of MIC members. The curriculum includes annotated reading lists, lecture notes, problem sets with solutions, and PennyLane/Qiskit implementation exercises. Designed for students with a classical ML background and no prior quantum experience.

EducationPennyLaneQiskitOpen SourceCurriculum

Leads: Aiden Park, Mei-Ling Chen

Contributors: Tobias Richter, Raj Patel, Yuki Tanaka

CompletedSince May 2023

Barren Plateau Diagnostic Toolkit

A lightweight Python package providing practical diagnostics for barren plateaus in parameterized quantum circuits. Implements gradient variance estimation, loss landscape visualization, and trainability scores to help researchers identify problematic circuits before committing to expensive hardware runs.

PythonPennyLaneOptimizationBarren PlateausTools

Leads: Priya Nair, Tobias Richter

ActiveSince Nov 2025

Hybrid Molecular Property Predictor

A hybrid quantum-classical architecture for predicting molecular properties relevant to drug discovery. Quantum circuit layers encode molecular symmetries via equivariant feature maps; classical transformer layers aggregate predictions across the molecular graph. Developed during the 2025 MIC Fall Hackathon and subsequently extended into an ongoing research project.

Hybrid SystemsDrug DiscoveryGraph Neural NetworksPennyLanePyTorch

Leads: Sofia Martinez, Nadia Hassan

Contributors: Amara Diallo, Leon Fischer

PlanningSince Jan 2026

MIC Knowledge Base

A community-maintained wiki covering quantum computing fundamentals, QML background reading, tutorial notes from past workshops, and internal research documentation. Hosted on GitHub Pages and openly accessible to the research community.

DocumentationEducationOpen Source

Leads: Yuki Tanaka, James Okonkwo

Collaborate

Partner with MIC on research

We actively seek partnerships with faculty, external research groups, and industry labs.

Faculty Collaboration

Propose a supervised project for MIC members, co-run a reading group, or offer mentorship in exchange for motivated research assistants.

Industry Partnership

Sponsor a project, provide cloud compute credits, or co-design a challenge problem. We produce rigorous, citable technical output.

External Researchers

Give a guest lecture, co-author a technical report, or open a collaborative research thread. We welcome connections across institutions.

Want to contribute to MIC research?

Join a stream, pick up an open GitHub issue, or pitch a new project. All levels of experience are welcome.