Parametric Neural Amp Modeling with Active Learning

2025-09-30

AI Summary

This work addresses the high data acquisition cost in parametric guitar amplifier modeling. We propose Panama, an end-to-end differentiable modeling framework that integrates LSTM with a WaveNet-inspired architecture, coupled with a novel ensemble-based active learning strategy that maximizes model disagreement via gradient-based optimization to dynamically select the most informative knob configurations. This significantly reduces the required number of measurements. Experiments show that Panama achieves audio fidelity comparable to the non-parametric benchmark NAM on MUSHRA subjective listening tests using only 75 parameter settings—substantially lowering hardware measurement overhead. The core contribution is the first integration of differentiable active learning with parametric neural audio modeling, enabling high-fidelity, low-sample-count, and interpretable speaker cabinet response modeling.

Technology Category

• Machine Learning: Active Learning
• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

We introduce Panama, an active learning framework to train parametric guitar amp models end-to-end using a combination of an LSTM model and a WaveNet-like architecture. With model, one can create a virtual amp by recording samples that are determined through an ensemble-based active learning strategy to minimize the amount of datapoints needed (i.e., amp knob settings). Our strategy uses gradient-based optimization to maximize the disagreement among ensemble models, in order to identify the most informative datapoints. MUSHRA listening tests reveal that, with 75 datapoints, our models are able to match the perceptual quality of NAM, the leading open-source non-parametric amp modeler.

Problem

• Develops active learning for parametric guitar amp modeling
• Minimizes required data points through ensemble disagreement strategy
• Matches perceptual quality of leading non-parametric models

Innovation

• Active learning framework trains parametric guitar amp models
• Combines LSTM and WaveNet-like architecture end-to-end
• Uses ensemble disagreement to minimize required data points