noa/thesis
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

very barebones introduction

Browse source
This commit is contained in:
Noa Aarts 2025-12-12 09:17:29 +01:00
parent 6b56ef7aff
commit bfd2584500
Signed by: noa
GPG key ID: 1850932741EFF672
1 changed files with 23 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

34
report/sections/1introduction.typ
View file

@ -5,15 +5,27 @@
= Introduction = Introduction
== Before starting In the NISQ era quantum-classical hybrid methods are better due to the limited depth, bla bla bla
Okay, so as far as I currently know a Variational Quantum Eigensolver (VQE) is a semi-quantum way to find eigenvalues
for a matrix. There has been significant research into various methods of Quantum Architecture Search (QAS),
They currently are mostly using Hardware Efficient Ansatze. // TODO: needs facts (list a bunch).
These Ansatze are optimized for relatively small connectivity, and don't keep errors into account that much. Most of these have been searches that are specific to the problem that is being solved,
I want to figure out if this can be done better, by using a descriptor format of the hardware including connections so the Paremetrized quantum circuits (PQCs) are generally not transferrable between different problems.
between qubits etc. is it possible to make ansatze that are better in
- fewer parameters There has also been a bit of research into task-agnostic QAS, where search
- less error prone can be performed once per hardware iteration instead of per problem, using proxies like expressibility
- speed of optimisation and entanglement (CITE THE PAPER).
Where maybe some finetuning can help specialise into a specific problem afterwards. (HYPOTHETICAL)
These task-agnostic searches have yet to include things like noise, arbitrary connectivity graphs
and allowed gate types on a per-qubit basis. And also haven't shown great scalability to architectures
with more qubits.
This research proposes a Task-Agnostic Evolutionary QAS implementing
hardware constraints and noise profiles able to target circuits with a
user-specified expressibility and entanglement to be able to mitigate the barren plateaus from
an expressibility that's too high (cite).
To achieve these goals a python library will be made that allows for composing various filters
and generators together (with multithreading) to help with developing and iterating on QAS methods.
That is what I will go and try figuring out I think