Dimitris G. Angelakis

Gordon Ma, Dimitris G. Angelakis

Qubit-efficient optimization seeks to represent an $N$-variable combinatorial problem within a Hilbert space smaller than $2^N$, using only as much quantum structure as the objective itself requires. Quadratic unconstrained binary optimization (QUBO) problems, for example, depend only on pairwise information -- expectations and correlations between binary variables -- yet standard quantum circuits explore exponentially large state spaces. We recast qubit-efficient optimization as a geometry problem: the minimal representation should match the $O(N^2)$ structure of quadratic objectives. The key insight is that the local-consistency problem -- ensuring that pairwise marginals correspond to a realizable global distribution -- coincides exactly with the Sherali-Adams level-2 polytope $\mathrm{SA}(2)$, the tightest convex relaxation expressible at the two-body level. Previous qubit-efficient approaches enforced this consistency only implicitly. Here we make it explicit: (a) anchoring learning to the $\mathrm{SA}(2)$ geometry, (b) projecting via a differentiable iterative-proportional-fitting (IPF) step, and (c) decoding through a maximum-entropy Gibbs sampler. This yields a logarithmic-width pipeline ($2\lceil\log_2 N\rceil + 2$ qubits) that is classically simulable yet achieves strong empirical performance. On Gset Max-Cut instances (N=800--2000), depth-2--3 circuits reach near-optimal ratios ($r^* \approx 0.99$), surpassing direct $\mathrm{SA}(2)$ baselines. The framework resolves the local-consistency gap by giving it a concrete convex geometry and a minimal differentiable projection, establishing a clean polyhedral baseline. Extending beyond $\mathrm{SA}(2)$ naturally leads to spectrahedral geometries, where curvature encodes global coherence and genuine quantum structure becomes necessary.

Jirawat Tangpanitanon, Chanatip Mangkang, Pradeep Bhadola et al.

Despite empirical successes of recurrent neural networks (RNNs) in natural language processing (NLP), theoretical understanding of RNNs is still limited due to intrinsically complex non-linear computations. We systematically analyze RNNs' behaviors in a ubiquitous NLP task, the sentiment analysis of movie reviews, via the mapping between a class of RNNs called recurrent arithmetic circuits (RACs) and a matrix product state (MPS). Using the von-Neumann entanglement entropy (EE) as a proxy for information propagation, we show that single-layer RACs possess a maximum information propagation capacity, reflected by the saturation of the EE. Enlarging the bond dimension beyond the EE saturation threshold does not increase model prediction accuracies, so a minimal model that best estimates the data statistics can be inferred. Although the saturated EE is smaller than the maximum EE allowed by the area law, our minimal model still achieves ~99% training accuracies in realistic sentiment analysis data sets. Thus, low EE is not a warrant against the adoption of single-layer RACs for NLP. Contrary to a common belief that long-range information propagation is the main source of RNNs' successes, we show that single-layer RACs harness high expressiveness from the subtle interplay between the information propagation and the word vector embeddings. Our work sheds light on the phenomenology of learning in RACs, and more generally on the explainability of RNNs for NLP, using tools from many-body quantum physics.

Beng Yee Gan, Daniel Leykam, Dimitris G. Angelakis

The data-embedding process is one of the bottlenecks of quantum machine learning, potentially negating any quantum speedups. In light of this, more effective data-encoding strategies are necessary. We propose a photonic-based bosonic data-encoding scheme that embeds classical data points using fewer encoding layers and circumventing the need for nonlinear optical components by mapping the data points into the high-dimensional Fock space. The expressive power of the circuit can be controlled via the number of input photons. Our work shed some light on the unique advantages offers by quantum photonics on the expressive power of quantum machine learning models. By leveraging the photon-number dependent expressive power, we propose three different noisy intermediate-scale quantum-compatible binary classification methods with different scaling of required resources suitable for different supervised classification tasks.