SCRAPL: Scattering Transform with Random Paths for Machine Learning

2026-02-11

AI Summary

This work addresses the computational inefficiency of conventional wavelet scattering transforms, whose prohibitively large number of paths hinders their integration as differentiable losses in neural network training. To overcome this limitation, the authors propose SCRAPL—an efficient optimization framework that leverages random path sampling, importance-sampling-based initialization, and differentiable digital signal processing (DDSP) to drastically reduce the computational cost of the joint time–frequency scattering transform (JTFS). Evaluated on unsupervised timbre matching tasks—such as resynthesizing sounds from granular synthesizers and emulating the iconic Roland TR-808 drum machine—the method significantly improves both training efficiency and perceptual audio quality. This represents the first practical application of JTFS in end-to-end differentiable audio synthesis.

Technology Category

• Search and Optimization: Mixed Discrete/Continuous Search
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Machine Learning: Time-Series/Data Streams

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

The Euclidean distance between wavelet scattering transform coefficients (known as paths) provides informative gradients for perceptual quality assessment of deep inverse problems in computer vision, speech, and audio processing. However, these transforms are computationally expensive when employed as differentiable loss functions for stochastic gradient descent due to their numerous paths, which significantly limits their use in neural network training. Against this problem, we propose"Scattering transform with Random Paths for machine Learning"(SCRAPL): a stochastic optimization scheme for efficient evaluation of multivariable scattering transforms. We implement SCRAPL for the joint time-frequency scattering transform (JTFS) which demodulates spectrotemporal patterns at multiple scales and rates, allowing a fine characterization of intermittent auditory textures. We apply SCRAPL to differentiable digital signal processing (DDSP), specifically, unsupervised sound matching of a granular synthesizer and the Roland TR-808 drum machine. We also propose an initialization heuristic based on importance sampling, which adapts SCRAPL to the perceptual content of the dataset, improving neural network convergence and evaluation performance. We make our code and audio samples available and provide SCRAPL as a Python package.

Problem

• Scattering Transform
• Computational Efficiency
• Differentiable Loss
• Neural Network Training
• Perceptual Quality Assessment

Innovation

• Scattering Transform
• Random Paths
• Differentiable Signal Processing
• Perceptual Quality Assessment
• Stochastic Optimization