Leonardo Lavagna

Leonardo Lavagna

I'm a mathematician specialized in Data Science currently pursuing a PhD in Quantum Computing

Semi-Supervised Multichannel Anomaly Detection Using QAOA-Driven Selection and Set Covering

less than 1 minute read

Published:

Our contribution on semi-supervised multichannel anomaly detection using QAOA and set covering, where multivariate time-series information is encoded into a QUBO and QAOA-driven samples are used to build interpretable convex regions of normal behavior, has been published in the 2026 International Joint Conference on Neural Networks (IJCNN). The method is evaluated on synthetic data and real-world water smart-meter measurements, achieving competitive performance and the best recall and F1-score among the considered methods on the Città Ducale dataset. The paper is available here. Code is available here.

You May Also Enjoy

From Problem-Structured QAOA to Problem-Agnostic Variational MaxCut Solvers: Representation, Redundancy, and Optimization Landscapes

less than 1 minute read

Published:

Our contribution on the relationship between QAOA layers and MaxCut structure via “agnostic” variational solvers where the problem structure is controllably removed and the corresponding solutions analyzed (in terms of cost landscapes and quality) has been published in Applied Mathematics and Statistics. The article is available here. Code available here and related blog post here. Read more

Papers related to the 2026 IEEE International Symposium on Circuits and Systems

less than 1 minute read

Published:

We recently presented two contributions at the 2026 IEEE International Symposium on Circuits and Systems (ISCAS) in Shanghai. These contributions are a natural continuation of our works about time-space trade-offs in quantum cryptosystems (I’ve talked about this research topic in a couple of blog posts, se e.g. here). In this case we focused on two complementary aspects of the quantum version of the “Hellman construction” proposed last year at the same conference in London (cf., this post and this paper)

  • A paper about a unified framework for cryptographic search based on quantum walks and Grover’s algorithm
  • A paper about the impossibility of having multiplicative resource scaling in cryptographic search under the previous unified model.
Read more

Enhancing QAOA Ansatz via Multi-Parameterized Layer and Blockwise Optimization

less than 1 minute read

Published:

The contribution about QAOA in the conference ICASSP 2026 (see this previous post) has been published here. I’ve also updated the blog, with a brief discussion about the main ideas behind the paper. In particular, in this paper we focus on multi-parameterized layers and advanced parameters’ update, in particular on the effect that an extra layer at the end of a standard QAOA circuit has, and on possible (blockwise) optimizations of the parameters of the extended circuit, including ablation studies. Concretely, applied to the MaxCut problem across diverse graph families, the proposed architecture achieves, with an order of magnitude smaller circuit depth, approximation ratios comparable to QAOA, using a single cost–mixer layer, thus reducing gate counts by up to fivefold. Moreover, ablation studies prove that blockwise fine-tuning is crucial to deliver higher-quality solutions at shallower depth, offering a practical, quantum hardware-efficient alternative for signal processing applications. Most of the code is an adaptation of this repository. Read more