Shot-Based Quantum Encoding Introduced for Efficient Data Loading in QML
A novel approach for data loading, termed shot-based quantum encoding (SBQE), has been introduced to overcome the challenges of efficient data loading in quantum machine learning for the near future. Current techniques like angle, amplitude, and basis encoding either fail to fully utilize the exponential capacity of Hilbert space or necessitate circuit depths that surpass the coherence limits of noisy intermediate-scale quantum systems. SBQE allocates the native resource of the hardware—shots—based on a classical distribution that depends on the data across various initial quantum states. By considering shot counts as a learnable parameter, SBQE yields a mixed-state representation with expectation values that are linear in classical probabilities, facilitating integration with nonlinear activation functions. This method mirrors a multilayer perceptron, with weights implemented through quantum circuits. A compatible implementation protocol is outlined. This research was published on arXiv under the identifier 2604.06135v2.
Key facts
- SBQE stands for shot-based quantum encoding.
- It addresses data loading inefficiency in near-term quantum machine learning.
- Existing encoding schemes include angle, amplitude, and basis encoding.
- SBQE uses a data-dependent classical distribution over initial quantum states.
- Shot counts are treated as a learnable degree of freedom.
- The method produces a mixed-state representation.
- SBQE is structurally equivalent to a multilayer perceptron with quantum circuit weights.
- The paper is on arXiv with ID 2604.06135v2.
Entities
Institutions
- arXiv