Quantum error correction using Deep learning

Pridgar, Blaž (author), Demšar, Jure (mentor), Granath, Mats (comentor)

Quantum computers hold promise for solving problems infeasible for classical machines. However, due to their susceptibility to errors, quantum error correction (QEC) is essential. This thesis investigates surface code decoders based on graph and recurrent neural networks, analyzing their predictions through soft outputs, uncertainty estimation, and hard case identification. For this purpose, we developed a pipeline for the analysis of models soft-outputs, revealing underlying properties of the data and behavior of models. We also established methods for detecting challenging syndromes and explored additional training on these cases. Furthermore, we also implemented a Bayesian approach to quantify uncertainty. Our results show that for lower code distances, decoders approximate maximum-likelihood decoding well, while the larger distance codes succumb to the sparsity of data. The hard case fine-tuning and the Bayesian approach did not significantly enhance performance of our models. Overall, our findings demonstrate the value of the analysis tools and training strategies developed in this work. These methods enabled us to better understand decoder behavior, identify sources of error, and outline concrete pathways for improving quantum error correction in future studies.

• Language: English
• Keywords: Quantum error correction, neural network decoders, graph neural networks, recurrent neural networks, soft-outputs
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2025
• PID: 20.500.12556/RUL-174273
• COBISS.SI-ID: 254383619
• Publication date in RUL: 30.09.2025

Secondary language

• Language: Slovenian
• Title: Kvantna korekcija napak z uporabo globokega učenja

Abstract

Kvantni računalniki obetajo reševanje trenutno nerešljivih problemov za klasične računalnike, vendar zaradi svoje občutljivosti na napake nujno potrebujejo metode kvantnega popravljanja napak (QEC). V magistrskem delu smo preučevali dekoderje površinskih kod, ki temeljijo na grafovskih in rekurenčnih nevronskih mrežah, ter analizirali njihove napovedi z vidika mehkih izhodov, ocene negotovosti in identifikacije težkih primerov. V ta namen smo razvili cevovod za analizo mehkih izhodov modelov, ki razkriva lastnosti podatkov in vedenje modelov. Poleg tega smo razvili metode za odkrivanje zahtevnih sindromov in preučili dodatno učenje na teh primerih. Uvedli smo tudi Bayesov pristop za kvantifikacijo negotovosti. Rezultati kažejo, da dekoderji pri manjših kodnih razdaljah dobro aproksi- mirajo maksimalno verjetnostno dekodiranje, medtem ko modeli z večjimi kodnimi razdaljami trpijo zaradi redkosti podatkov. Dodatno učenje na težkih primerih in Bayesov pristop nista bistveno izboljšala zmogljivosti modelov. Naše ugotovitve potrjujejo uporabnost razvite analitične metode in strategij učenja, saj omogočajo boljše razumevanje vedenja dekoderjev, prepoznavanje virov napak in načrtovanje prihodnjih izboljšav kvantnega popravljanja napak.

• Keywords: Kvantno popravljanje napak, dekoderji iz nevronskih mrež, grafovske nevronske mreže, rekurenčne nevronske mreže, mehki izhodi