Improving Fine-Tuning with Latent Cluster Correction
This addresses a specific issue in fine-tuning for classification tasks, but it appears incremental as it builds on existing clustering concepts.
The paper tackled the problem of improving fine-tuning performance by optimizing latent cluster formation in neural networks, achieving preliminary results on CIFAR-100 that demonstrate the method's viability.
The existence of salient semantic clusters in the latent spaces of a neural network during training strongly correlates its final accuracy on classification tasks. This paper proposes a novel fine-tuning method that boosts performance by optimising the formation of these latent clusters, using the Louvain community detection algorithm and a specifically designed clustering loss function. We present preliminary results that demonstrate the viability of this process on classical neural network architectures during fine-tuning on the CIFAR-100 dataset.