Spirals Dataset

Hard

Two intertwined spirals - the ultimate quantum ML challenge

Dataset Overview

Key characteristics and specifications

Samples

300

Features

2 (X, Y coordinates)

Classes

2 (Binary)

Difficulty

Hard

The Spirals dataset is one of the most challenging binary classification problems. Two intertwined spirals rotate around a common center with different angular velocities. This dataset severely tests the expressiveness of any machine learning model and is an excellent testbed for demonstrating quantum advantage in learning complex, highly non-linear patterns.

Dataset Visualization

Spirals dataset showing two intertwined spirals

Quantum Advantage

Why quantum is needed for spirals

Classical Limitation

Classical neural networks need exponentially many hidden units to separate spirals, while quantum circuits can achieve this with polynomial resources using quantum superposition and entanglement.

where n = feature dimension

Deep Variational Circuit

Advanced quantum architecture for complex patterns

Multi-Layer Ansatz

Deep quantum circuits with multiple entangling layers to learn intricate spiral boundaries

where:

Spiral Geometry

Mathematical properties of spiral patterns

Each spiral can be parameterized in polar coordinates as a curve where the distance from origin increases with angle.

Spiral parameterization:

Cartesian conversion:

Training Characteristics

Expected convergence behavior

Epochs Recommended

500-1000 (significantly more than simpler datasets)

Learning Rate

0.01 - 0.05 (may require careful tuning)

Convergence Time

Slower than other datasets - barren plateaus possible