Few-shot Class-variable Incremental Audio Classification via Prototype Adaptation and Pseudo Class-variable Training

2026-06-08 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

A new method addresses Few-shot Class-variable Incremental Audio Classification (FCIAC), a problem where the number of audio classes can dynamically increase or decrease, departing from traditional few-shot class-incremental learning that only accounts for class additions. The proposed FCIAC method integrates prototype adaptation and pseudo class-variable training. Its model comprises an encoder and a classifier, with the classifier initialized by a class-variable prototype adaptation network designed to dynamically adjust its structure based on class changes. Furthermore, a pseudo class-variable training strategy is employed to bolster the model's adaptability to these fluctuating class numbers. Experimental results across three public datasets demonstrate that this method achieves superior average accuracy compared to existing approaches. The code is publicly available on GitHub.

Key takeaway

For Machine Learning Engineers developing audio classification systems where class numbers fluctuate, you should consider implementing dynamic prototype adaptation and pseudo class-variable training. This approach, demonstrated to improve average accuracy on public datasets, offers a robust solution for real-world scenarios beyond simple class additions. Integrating this method can significantly enhance your model's adaptability and performance in variable class environments.

Key insights

The paper introduces FCIAC, a dynamic audio classification problem, solved with prototype adaptation and pseudo class-variable training for improved accuracy.

Principles

• Class numbers can both increase and decrease.
• Dynamic network structures enhance adaptability.
• Pseudo-training improves model robustness.

Method

The method uses an encoder and a classifier initialized by a dynamically changing class-variable prototype adaptation network. A pseudo class-variable training strategy further enhances adaptability to varying class counts.

In practice

• Apply dynamic class adaptation in audio systems.
• Implement pseudo-training for fluctuating datasets.
• Use prototype adaptation for few-shot learning.

Topics

• Few-shot Learning
• Audio Classification
• Incremental Learning
• Prototype Adaptation
• Class-variable Training
• Deep Learning

Code references

• cgq2971-afk/FCIAC

Best for: Research Scientist, AI Scientist, Machine Learning Engineer

Related on AIssential

• Incremental learning for audio classification with Hebbian Deep Neural Networks — Hebbian learning with kernel plasticity improves incremental audio classification by balancing new learning and knowledge retention.
• Learning Bayesian Network Classifiers to Minimize Class Variable Parameters — A new Bayesian network classifier minimizes class variable parameters while asymptotically estimating true probability distributions.
• ADAPTive Input Training for Many-to-One Pre-Training on Time-Series Classification — ADAPT enables mixed-batch pre-training for time-series data by aligning physical properties across diverse datasets.
• Open World Autoencoding Drift Detection with Novel Class Recognition in Tabular Non-stationary Data Streams — An unsupervised method detects concept drift via autoencoder reconstruction errors and novel classes using density estimation on proxy representations.
• Re-Evaluating Continual Learning with Few-Shot Adaptation — Few-shot evaluation offers a more comprehensive assessment of continual learning model stability and plasticity than traditional 0-shot methods.
• Stage-adaptive audio diffusion modeling — Optimizing audio diffusion training requires dynamic adaptation to evolving semantic and refinement priorities.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.