Page 72 - Kaleidoscope Academic Conference Proceedings 2024

2024 ITU Kaleidoscope Academic Conference

states. Next, QSVC [23] chooses a certain hyperplane in a
quantum feature space, which aims to minimize classification
errors and obtain the greatest distance between classes. Thus,
we may formulate it as a quadratic programming problem
where we seek to minimize the objective function given the
established constraints.

4.3 Quantum K-Mean Clustering

Quantum K-Mean Clustering (QKMC) [20] enables the laws
of quantum mechanics to perform clustering operations. In
QKMC, the data vectors are mapped to quantum states by
phases using a circuit of quantum interconnections, with
each data point being represented by a quantum state in a
Figure 4 – Major QML Techniques Classification. higher-dimensional Hilbert space. This wiring is conveyed
mathematically as | ⟩ = (x )|0⟩, where x represents the
As far as mathematically QML might use the quantum
-th data set, and (x ) is a quantum circuit that encodes data
computer’s unique characteristics, which are superposition,
x into states. The QKMC teams carry out the process of
entanglement, and interference in overcoming the classic
quantum operations periodically to group the quantum states
learning tasks more efficiently. With implementation of
into clusters, but the distance between the intra-states and
quantum algorithms and quantum data representations in
distance between the inter-states have to be minimized and
QML, it targets to address hard problems of this domain,
maximized respectively. In QKMC, the optimization process
which include optimization, pattern recognition and data
is highly complicated, where the position of the centroids of
analysis. The primary purpose of QML is to open up
clusters occurs by adjusting the parameters of the quantum
the marine of mind-blowing technologies that quantum
circuit in order to minimize a distance function.
computing application might contribute to enhancing the
efficiency and precision of machine learning systems.
5. PERFORMANCE EVALUATION AND
IMPLEMENTATION
4.1 Variational Quantum Classifier (VQC)
In classification problems, performance evaluation metrics
VQC [18] is a QML algorithm that uses regularized PQCs
(parameterized quantum circuits) to classify inputs. Let a set [21] are essential for defining the success of predictive models
by the fact of dividing data into previously set classes. Metrics
of input features denoted by x and the corresponding class
which are usually used comprise accuracy, precision, recall,
labels given by y, the VQC encodes the single feature into
and F1 score.
a quantum state with the trainable system parameters via
parameterized quantum circuit ( ). It [22] encodes data
in classical form into quantum representations that are used +
Accuracy = (6)
next for quantum computations. Lastly, a quantum measuring + + +
procedure is performed, generating results that are used as
classifiers. The optimization of the parameters is done

to get a minimal cost function ( ), which characterizes Precision( ) = (7)
+
the disparity between the true labels and the predicted
ones. Generally speaking, neural networks employ classical
optimization algorithms like gradient descent. Quantum Precision( ) = (8)
computers are powerful machines that can be trained to solve +
classification problems characterized by a large number of
measurements.
Sensitivity = Recall( ) = (9)
+
4.2 Quantum Support Vector Classifier (QSVC)

The QSVC [19] is a method that is explicit in the following Specificity = Recall( ) = (10)
way: the quantum kernel function (x , x ) is defined as +
a quantum circuit (x , x ) mapping input data points x
and x into the higher-dimensional quantum feature space. 2 · Precision( ) · Recall( )
F-Measure = F-1 Score =
Allegorically, this transformation can be designated as (x ) Precision( ) + Recall( )
and (x ) mathematically. In this quantum feature space, the (11)
inner product ⟨ (x ), (x )⟩ is actually the quantum kernel The Methodology followed in this paper is written in the
(x , x ), which is responsible for expressing the similarity below Algorithm, which indicate the several steps for the
between input points in terms of their corresponding quantum Classification of PIMA Diabetes Dataset using ML and QML

– 28 –