Class-Incremental Mixture of Gaussians for Deep Continual Learning
This work addresses the problem of flexible and efficient continual learning for AI systems that need to adapt to new classes over time, representing an incremental improvement over existing centroid-driven methods.
The paper tackles the challenge of class-incremental continual learning, where classes arrive sequentially without grouping, by proposing an end-to-end mixture of Gaussians model integrated with a deep feature extractor. The result is a competitive method that effectively learns in memory-free scenarios and demonstrates strong performance compared to state-of-the-art baselines in image classification.
Continual learning models for stationary data focus on learning and retaining concepts coming to them in a sequential manner. In the most generic class-incremental environment, we have to be ready to deal with classes coming one by one, without any higher-level grouping. This requirement invalidates many previously proposed methods and forces researchers to look for more flexible alternative approaches. In this work, we follow the idea of centroid-driven methods and propose end-to-end incorporation of the mixture of Gaussians model into the continual learning framework. By employing the gradient-based approach and designing losses capable of learning discriminative features while avoiding degenerate solutions, we successfully combine the mixture model with a deep feature extractor allowing for joint optimization and adjustments in the latent space. Additionally, we show that our model can effectively learn in memory-free scenarios with fixed extractors. In the conducted experiments, we empirically demonstrate the effectiveness of the proposed solutions and exhibit the competitiveness of our model when compared with state-of-the-art continual learning baselines evaluated in the context of image classification problems.