Banca de DEFESA: LUCIANO SERAFIM DE SOUZA

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : LUCIANO SERAFIM DE SOUZA
DATE: 27/08/2024
TIME: 09:00
LOCAL: https://meet.google.com/fns-khjw-tzp
TITLE:

Lackadaisical Quantum Walk Analyses With Partial Phase Inversion Proposal

KEY WORDS:

Quantum Computing. Artificial Neural Networks Training.  Quantum Walk.  Quantum Search Algorithm.  Lackadaisical Quantum Walk. Multiple Self-loops. Partial Phase Inversion. Adjacent Marked Vertices.

PAGES: 100
BIG AREA: Ciências Agrárias
AREA: Agronomia
SUMMARY:

Quantum walks have been a theme of much interest in the field of quantum computing. Defined as a theoretical model that describes the movement of a particle in a discretized  or continuous space of time, they apply the principles of quantum mechanics and present properties with great practical potential. The development of quantum search algorithms represents the main application of quantum walks. These search algorithms based on quantum walks have shown several promising results providing efficient solutions to complex problems. However, some theoretical and practical challenges still need to be overcome. In this thesis work, we use the quantum walks defined in a discrete-time space. Initially, we apply a quantum walk on the complete graph to develop a quantum search procedure capable of finding a set of synaptic weights that train a classical artificial neural network. The quantum walk in the complete graph needs a quantum operator to perform a transformation from an n-dimensional space to a four-dimensional space, which has not yet been defined. Furthermore, as the vertices of the graph are indistinguishable, their locations are given by a mapping on an n-dimensional grid. In this way, we decided to analyze the lackadaisical quantum walk in the hypercube since it is possible to reduce it to a line walk. Based on this analysis, we propose a weight value for the self-loop, which is ideal for searching for multiple marked vertices. It was possible to achieve maximum success probabilities close to 1. The results also show that adjacent marked vertices decrease the maximum probability of success. So we propose a new approach that uses multiple self-loops at each vertex and applies the partial inversion of the target state called Multi-self-loop Lackadaisical Quantum Walk. With this new approach, we also propose two new weights and achieve maximum success probabilities close to 1, even in cases where there are adjacent marked vertices.

COMMITTEE MEMBERS:
Externo à Instituição - FRANKLIN DE LIMA MARQUEZINO - UFRJ
Externo à Instituição - ADENILTON JOSE DA SILVA - UFPE
Externo ao Programa - 2258157 - JOSE FERRAZ DE MOURA NUNES FILHO - nullPresidente - TIAGO ALESSANDRO ESPINOLA FERREIRA
Interno - WILSON ROSA DE OLIVEIRA JUNIOR