Search papers, labs, and topics across Lattice.
This paper introduces the Dynamic Neural Graph Encoder (DNG-Encoder), which utilizes dynamic graphs to effectively represent and process the high-dimensional weight spaces of neural networks while preserving the sequential nature of inference. By leveraging this framework, the authors develop INR2JLS, a method for classifying Implicit Neural Representations (INRs) that significantly enhances performance. The results show a notable improvement of approximately 10% in INR classification accuracy on the CIFAR-100-INR dataset compared to existing state-of-the-art methods.
Dynamic graphs can unlock a 10% boost in classification accuracy for Implicit Neural Representations, redefining how we approach neural network weight spaces.
The rapid advancements in using neural networks as implicit data representations have attracted significant interest in developing machine learning methods that analyze and process the weight spaces of other neural networks. However, efficiently handling these highdimensional weight spaces remains challenging. Existing methods often overlook the sequential nature of layer-by-layer processing in neural network inference. In this work, we propose a novel approach using dynamic graphs to represent neural network parameters, capturing the temporal dynamics of inference. Our Dynamic Neural Graph Encoder (DNG-Encoder) processes these graphs, preserving the sequential nature of neural processing. Additionally, we also leverage DNG-Encoder to develop INR2JLS (Implicit Neural Representation to Joint Latent Space) for facilitate downstream applications, such as classifying Implicit Neural Representations (INRs). Our approach demonstrates significant improvements across multiple tasks, surpassing the state-of-the-art INR classification accuracy by approximately 10% on the CIFAR-100-INR.
