<ul><li>Learning curve extrapolation predicts neural network performance from early training epochs.</li><li>Existing methods neglect the impact of neural network architectures on learning curves.</li><li>A novel architecture-aware neural differential equation model is proposed to forecast learning curves continuously.</li><li>The model outperforms current state-of-the-art methods and pure time-series modeling approaches.</li></ul>

Architecture-Aware Learning Curve Extrapolation via Graph Ordinary Differential Equation

Discover more