Gaussian Lowpass Dataset#

These datasets contain sequences of quantized Gaussian noise that has been smoothed using a moving average (lowpass) filter. The data is generated by:

1. Drawing samples from a Gaussian distribution (σ=5)
2. Applying a moving average filter of varying window sizes
3. Quantizing the filtered result to integers

The moving average filter acts as a lowpass filter, attenuating high-frequency components while preserving low-frequency trends. Larger window sizes result in smoother sequences with stronger low-frequency characteristics.

Variants#

We provide three variants with different window sizes:

• gaussian-lp-03 (window=3): Light smoothing
• gaussian-lp-07 (window=7): Moderate smoothing
• gaussian-lp-15 (window=15): Strong smoothing

Each variant uses σ=5 for the initial Gaussian noise and applies integer quantization after filtering to maintain compatibility with integer-based compression algorithms.