Quantum Physics Letters
Journal license

Journal

Quantum Physics Letters


Volume
& Issue

Volume 6, Issue 2


Published
on

August 1, 2017


Pages

95-98


DOI

Article

A Simple Quantum Algorithm for Exponentially Fast Target Searching in the Unstructured Database

Check for updates


Authors

Dhananjay P. Mehendale* Affiliation:
Department of Physics, Savitribai Phule University of Pune, Pune, India-411007


Abstract

We present an exponentially fast simple quantum algorithm for searching the unknown target in the unstructured database. This algorithm provides an eloquent example that clearly demonstrates the enormous advantage of quantum parallelism. The main idea behind achieving exponential speedup for this new quantum algorithm over Grover’s quantum algorithm is actually very simple. The idea consists of simultaneously employing $(n/2)$ oracles or black-box functions instead of utilizing only one oracle or black-box function as is done by Grover’s quantum algorithm for searching the target in the unordered data set of size $N = 2n$. We show that we can attain the (explicitly unknown) target in the unstructured database of size $N = 2n$ by giving in parallel only one call, simultaneously and independently, to appropriately defined $(n/2)$ oracles or black-box functions to be implemented using a quantum computer. The essential idea is to decompose the operation to be done on entire quantum system into $(n/2)$ operations to be carried out in parallel, simultaneously and separately, on individual components of the system and thus to achieve enormous speedup in obtaining the desired target from the unstructured data set of size $N = 2n$ which is indeed amazing.


Keywords

Unstructured database search, Quantum parallelism, Multiple number of Oracles


Citation

Mehendale, D. P. (2017). A simple quantum algorithm for exponentially fast target searching in the unstructured database. Quantum Physics Letters, 6(2), 95–98. https://doi.org/10.18576/qpl/060203

Published by: Engineering Journals

Engineering Journals Logo