Skip to content

Data Encoding and Kernel Notes

Andrew Musselman edited this page Sep 20, 2024 · 1 revision
  1. Dataset Prep: Use an existing toy binary classification dataset from this qiskit example. The CSV is generated here in my kernel dev branch if you’d like to glance at it.
    1. Note: this data is generated using the ad_hoc_data function in the qiskit-machine-learning library
  2. Data Encoding: recreate "from scratch" the effect of ZZFeatureMap on the dataset, but using circuits with a Cirq backend
    1. ZZFeatureMap is a ubiquitous example feature map used in Qiskit, and it utilizes a special implementation of angle encoding. First introduced in this paper
  3. Kernel Calculation: recreate "from scratch" the effect of FidelityKernel, using circuits with a Cirq backend
  4. Validation: Return an identical outcome to the Qiskit SVM example using pieces constructed in the preceding steps
  5. With the current learnings in hand, ideate possible designs which allow a from-scratch implementation of the necessary data encoding and kernel calculation in a backend-agnostic way (at the very least between Qiskit and Cirq)
Clone this wiki locally