Sparsity and Symbols with Kolmogorov-Arnold Networks

Explore Kolmogorov-Arnold Networks (KANs) through this mathematical lecture that examines their theoretical foundations and practical applications in sparsity and symbolic regression. Review essential results regarding KANs and discover how sparsity masks create connections between deep neural networks and Kolmogorov-Arnold Networks. Learn how KANs can be effectively combined with multimodal language models to perform symbolic regression tasks. Gain insights into both theoretical developments and empirical findings through a presentation that emphasizes mathematical rigor with chalk-based explanations supplemented by key empirical results.