Analyzing the effect of feature mapping techniques along with the circuit depth in quantum supervised learning by utilizing quantum support vector machine

DEQSVC: Dimensionality Reduction and Encoding Technique for Quantum Support Vector Classifier Approach to Detect DDoS Attacks

“…To overcome this issue, we used the fast ICA Dimensionality reduction technique to reduce the dimensions of the training dataset to two dimensions. The fast ICA is an enhanced version of the ICA and is a common dimensionality technique that is used to separate variables that have more than one outcome into independent sub-components [25].…”

“…The goal is to transform x using a linear static transformation into a vector of maximally sub-independent components estimated by some function 𝐹(𝑠 , 𝑠 , … . , 𝑠 ) of independence [25].…”

“…As seen in Figure 2, the fast ICA process generates two features, and we have only two labels: Benign and DDoS_LDAP. We first applied the PCA, but it causes DEQSVC to generate poor detection accuracy [25]. While the ICA looks for independent variables, the PCA technique only searches for uncorrelated relationships.…”

“…And according to equation equation (19), and [30], 𝑙 𝑤; (𝜙(𝑥), 𝑦) is represented by the following formula: 𝑙 𝑤; (𝜙(𝑥), 𝑦) = max{0,1 − 𝑦〈𝑤, 𝜙(𝑥)〉} (25) Recall that the data mapping function can be estimated using kernel evaluations, although for sake of simplicity we continue to use the data mapping representation 𝜙(𝑥) when we refer to kernel estimation. Shalev-Shwartz proves for all 𝑡 we can rewrite 𝑤 as:…”