Concepts¶
Understanding quantum data encoding requires bridging classical machine learning with quantum information theory. This section provides the conceptual foundations you need to use the Quantum Encoding Atlas effectively.
What You'll Learn¶
-
Why classical data must be transformed before a quantum computer can process it, and the different strategies for doing so.
-
The metrics that characterise an encoding: expressibility, entanglement capability, trainability, circuit depth, and simulability.
-
Why all-pairs testing has zero power across a handful of datasets, and the Friedman/Nemenyi procedure that replaces it — including what a missing result must not be imputed as.
-
Why a metric that is structurally zero and one that was never measured both arrive as
None, demand opposite treatment, and why imputing a column median for the first manufactured a robustness claim. -
When and why quantum encodings outperform classical feature maps, and what theoretical guarantees exist.
The Big Picture¶
Classical Data Quantum Encoding Quantum Processing
┌──────────────┐ ┌──────────────────┐ ┌──────────────────┐
│ │ │ │ │ │
│ x = [x₁, │ ────► │ |ψ(x)⟩ = U(x) │ ────► │ Measurement, │
│ x₂, │ Encode │ |0⟩⊗ⁿ │ │ Kernel, or │
│ ...] │ │ │ │ Variational │
│ │ │ │ │ Optimization │
└──────────────┘ └──────────────────┘ └──────────────────┘
The encoding step is the bridge. Its design determines what the
quantum computer can learn from your data.
Choosing an encoding is one of the most consequential decisions in a quantum machine learning pipeline. A poor choice can make quantum processing no better than classical — or even worse. A good choice can unlock computational structures that no classical kernel can efficiently replicate.
The pages in this section give you the vocabulary and intuition to make that choice deliberately.