Quantum
Machine Learning Conference 2026

• Date: 27.06.2026

• Venue: Online event (Zoom)

Paweł Gora

"Introduction to Quantum Machine Learning"

Abstract: In this talk, I will give an introduction to quantum machine learning. First, I will explain what quantum computing is and how it is different from classical computing paradigm. Later, I will tell how machine learning could be enhanced using quantum computing, presenting some of the possible opportunities and existing challenges. Finally, I will present how machine learning could be applied in the quantum computing domain, and outline how one can continue education and start a professional career in the quantum machine learning domain. 

Sven Groppe

"Data Quality in the Era of Quantum Computing"

Abstract: In an era in which data underpin decision-making across science, politics, and economics, ensuring high data quality is paramount. Conventional computing algorithms for enhancing data quality, including anomaly detection, demand substantial computational resources, lengthy processing times, and extensive training datasets. This work surveys the potential advantages of quantum computing for improving data quality, with a particular focus on anomaly detection. We begin by examining quantum techniques that could replace key subroutines in conventional anomaly-detection frameworks to mitigate their computational cost. This survey will especially serve as a guide to see the connection between classical and quantum methods in the data quality domain. We identify unresolved challenges and limitations in applying quantum computing to data quality tasks. Our findings open new avenues for research and applications that aim to enhance data-quality solutions using quantum technologies. 

Artur Miroszewski

"Adaptive Measurement Allocation for Learning Kernelized SVMs Under Noisy Observations"

Abstract:  Kernel methods are typically formulated under the assumption of exact, noise-free access to the Gram matrix. However, in emerging settings such as quantum machine learning, each kernel entry must be inferred from noisy observations, and its accuracy depends on how a limited measurement budget is allocated. Despite this, existing approaches overwhelmingly rely on uniform allocation, which equalizes estimator variance but ignores the highly non-uniform dependence of kernelized classifiers on the Gram matrix.

In this work, we introduce an adaptive measurement-allocation strategy for learning kernelized Support Vector Machines (SVMs) from noisy Bernoulli observations. Our approach combines two complementary principles: (i) geometric sensitivity, capturing how perturbations of individual kernel entries affect the classifier margin, and (ii) active-set instability, quantifying the probability of discrete changes in support-vector membership induced by measurement noise. These signals define a task-aware allocation scheme that concentrates measurements on the most decision-critical regions of the kernel matrix.

We provide a theoretical analysis showing that the benefit of adaptive allocation is governed by the heterogeneity of the induced kernel importance structure, leading to distinct regimes in which adaptive or uniform strategies are preferable. Empirical evaluations on synthetic datasets demonstrate that adaptive allocation significantly improves support-vector recovery, margin estimation, and decision-function accuracy under fixed measurement budgets. A dual-coefficient stability criterion further enables early stopping, achieving near-optimal performance while using only a fraction of the measurement cost. Additional experiments on quantum kernels derived from real-world data reveal a regime-dependent behavior aligned with known phenomena such as kernel concentration.

Together, these results establish adaptive measurement allocation as an effective alternative to uniform sampling for learning with noisy kernels, offering improvements in classifier fidelity and overall computational efficiency.

Jacob Cybulski

"Paradoxes of High-Dimensional Quantum Models"

Abstract: This presentation explores the intricate training dynamics of quantum machine learning models, framed as an interplay between classical optimization and the vast geometry of the Hilbert space. We move beyond the common trope of Hilbert space “vastness,” focusing instead on a geometry sculpted by quantum entanglement and operational constraints resulting in counter-intuitive paradoxes of high-dimensional models. We elucidate the classical optimizer’s struggle to refine classical parameters without insight into their quantum representation or their role in evolving qubit states. From a purely geometric perspective, we analyze how the high dimensionality of both the parameter and Hilbert spaces culminates in training pathologies like Barren Plateaus, Orthonormal Desert, and the Measurement Sparsity. Ultimately, we argue for the necessity of “quantum-aware” optimizers—tools capable of navigating the subtle curvature of the quantum manifold that standard gradient-based methods ignore. Keywords: Quantum Machine Learning, Quantum Model Optimization, Hilbert Space, Curse of Dimensionality, Barren Plateaus, Orthonormal Desert, Measurement Sparsity, Quantum Natural Gradient Reference: Jacob L. Cybulski, “A Dance of the Blind Puppeteer: The Interplay Between a Classical Optimizer and the Hilbert Space”, in Oswaldo Zapata (Editor), A Portrait of Quantum Technologies in Finance, The Quantum Finance Boardroom, 2026. 

Sebastian Zając

"Does Your Feature Map Actually Reach Quantum Space?"

Abstract: The feature map — the function that embeds classical data into a quantum state — is the primary design decision in any variational quantum classifier. Yet not every map into Hilbert space is genuinely quantum: some encodings, however natural they appear, produce hypothesis classes that are provably equivalent to classical kernel machines, regardless of ansatz depth or entanglement. In this talk we identify the precise mathematical boundary between classical and quantum encodings, grounded in information geometry and the algebra of observable transformations. We show that this boundary is both computable before training and directly linked to the training pathologies — barren plateaux, metric mismatch, vanishing gradients — that arise when a classical optimiser navigates a manifold it cannot see. We present a concise set of encoding diagnostics that allow practitioners to audit a feature map in advance, predict whether genuine quantum advantage is reachable, and choose encodings that actually exploit what the quantum hardware provides. 

Sebastian Dziura

"Multi-Source Classification with Quantum Architecture Search"

Abstract: With fault-tolerant quantum computing on the horizon, there is growing interest in applying quantum computational methods to data-intensive scientific fields like remote sensing. Quantum machine learning (QML) has already demonstrated potential for such demanding tasks. One area of particular focus is quantum data fusion—a complex data analysis problem that has attracted significant recent attention. In this work, we introduce an automated QML (AQML) approach for addressing data fusion challenges. We evaluate how AQML-generated quantum circuits perform compared to classical multilayer perceptrons (MLPs) and manually designed QML models when processing multisource inputs. Furthermore, we apply our method to change detection using the multispectral ONERA dataset, achieving improved accuracy over previously reported QML-based

change detection results. 

Karol Bartkiewicz

"Quantum Associative Memory with Photonic Quantum Memristors"

Abstract: Classical memristors — resistive devices with memory — are the key building blocks of neuromorphic hardware for associative memory, most notably Hopfield networks. In this talk, I will present a theoretical framework for a photonic quantum memristor that operates at the single-photon level and could serve as the elementary unit of a quantum associative memory. The device implements memory-dependent unitary transformations on polarization-encoded qubits via measurement-based feedback: a half-wave plate whose rotation angle evolves as dθ/dt = η N(t) − γ(θ − θ0), driven by photon detection events N(t). I will show that eliminating the classical memory variable from the coupled quantum–classical dynamics yields a non-Markovian integro-differential master equation with an explicit memory kernel within the Nakajima–Zwanzig framework, and that the hysteresis loop area is directly related to the kernel’s spectral density. This provides a rigorous bridge between memristive physics and open quantum system theory. I will then discuss how networks of such quantum memristors can implement quantum Hopfield networks, where memory-dependent coupling strengths and local fields enable history-dependent pattern storage and retrieval in the quantum regime. Crucially, the choice of feedback measurement basis dramatically alters the hysteresis characteristics — a uniquely quantum signature with no classical analogue, quantified by the measurement-basis sensitivity index and quantum hysteresis witness introduced in our work. I will present numerical results from stochastic quantum trajectory simulations validating the analytical predictions and outline a concrete experimental proposal using heralded single photons, feedback-controlled wave plates, and single-photon detectors. Finally, I will discuss how the inherent non-Markovian memory of quantum memristors offers advantages for temporal sequence processing in quantum neural networks, including recurrent architectures that encode temporal correlations without requiring circuit depth proportional to the sequence length.