Data-Reuploading Classifier

Data-Reuploading Classifier

This PennyLane demo implements the data-reuploading classifier, an elegant quantum machine-learning model that shows a single qubit can be a universal classifier when classical data is fed into the circuit multiple times. The key idea, introduced by Perez-Salinas and collaborators, is to interleave layers that re-encode the input data with layers of trainable rotations, so the data is uploaded repeatedly throughout the circuit rather than only once at the start. This repeated reuploading lets even a single-qubit circuit represent highly nonlinear decision boundaries, drawing an analogy to how repeated nonlinearities give classical neural networks their expressive power. The tutorial builds the alternating data-encoding and trainable layers in PennyLane, defines a fidelity-based cost that pushes data points toward target states on the Bloch sphere for each class, and trains the model with gradient-based optimization to classify non-trivially-shaped datasets. It visualizes the resulting decision regions and discusses how adding qubits and layers increases capacity. By demonstrating universal classification from minimal quantum resources, the demo is a striking, foundational example of expressive quantum machine learning in PennyLane.

Run it

pip install -r requirements.txt
python demo.py

Source and license

Imported from demonstrations_v2/tutorial_data_reuploading_classifier/demo.py in PennyLaneAI/demos at c52c0abeb5122218aa96b38eea848864cce7323f, under the Apache License 2.0. Original authors: Xanadu and the PennyLane community. The upstream LICENSE is included alongside this example.