Papers

M Bilkis, M Cerezo, G Verdon, PJ Coles, L Cincio · 2021 · 60 cites
A variable structure approach that grows and removes gates to keep quantum circuits shallow. Mitigates trainability and noise issues in variational algorithms. Demonstrated on VQE, quantum autoencoders and unitary compilation.
M Bilkis, M Rosati, RM Yepes, J Calsamiglia · Phys. Rev. Research, 2020 · 22 cites
Reinforcement learning agent that calibrates adaptive quantum receivers from scratch. Learns near-optimal setups under trial-and-error episodes. Benchmarked against Helstrom bound and Kennedy/Dolinar receivers.
M Bilkis, M Rosati, J Calsamiglia · IEEE ITW, 2021 · 10 cites
Extends RL-based receiver calibration to fading channels with variable transmissivity. Shallow RL methods achieve near-optimal performance. Presented at the IEEE Information Theory Workshop.
G Gasbarri, M Bilkis, E Roda-Salichs, J Calsamiglia · Quantum 8, 2024
Sequential analysis of streaming measurement data for quantum hypothesis testing. Stopping-time strategies outperform fixed-time approaches. Resource savings by a factor of 3-4 in optomechanical case studies.
M Bilkis, N Canosa, R Rossignoli, N Gigena · Phys. Rev. A, 2019
Analysis of conditional states and quantum correlations in hybrid qudit-qubit systems. Explores entropic measures of quantum discord. Published as part of the licenciatura thesis work at UNLP.
T Crosta, L Rebón, F Vilariño, JM Matera, M Bilkis · 2024
RL-based automatic recalibration for drifting quantum devices. Supervised master thesis work awarded Mención Masperi AFA 2024. Best master thesis in Physics in Argentina.
Y Cordero, S Biswas, F Vilariño, M Bilkis · 2024
Vectorized sketch representations combined with hybrid classical/quantum models. Benchmarked on QuickDraw sketch recognition with small-qubit circuits. Supervised master thesis at CVC-UAB.
M Bilkis, JM Kohler, F Vilariño · 2024
A multi-disciplinary teaching methodology for AI students combining device-building with social impact analysis. Piloted at UAB with strong student reception. Published methodology for EDULEARN24.
M Delgado, M Llopart, E Sarabia, S Taboada, P Vierge, F Vilariño, ... · 2024
System linking Spotify songs to clothing pattern generation via color palette extraction. Uses k-means clustering and Rhino/Grasshopper for design. Part of the UAB-Cruilla Chair on AI and music.

Theses

PhD thesis · Universitat Autònoma de Barcelona · 2023 · Cum Laude
Reinforcement learning methods for quantum control, from model-free to model-aware approaches. Directed by Prof. John Calsamiglia. Defended with grade 10/10 and Cum Laude distinction.
Licenciatura thesis · Universidad Nacional de La Plata · 2018
Conditional entropy and quantum correlations in hybrid qudit-qubit systems. Directed by Prof. Raúl Rossignoli and Prof. Norma Canosa. Defended with grade 10/10, graduated with honours.