Improved Dynamic Colouring of Sparse Graphs

Abstract: Given a dynamic graph subject to edge insertions and deletions, we show how to update an implicit representation of a proper vertex colouring, such that colours of vertices are computable upon query time. We give a deterministic algorithm that uses $O(\alpha^2)$ colours for a dynamic graph of arboricity $\alpha$, and a randomised algorithm that uses $O(\min{\alpha \log \alpha, \alpha \log \log \log n})$ colours in the oblivious adversary model. Our deterministic algorithm has update- and query times polynomial in $\alpha$ and $\log n$, and our randomised algorithm has amortised update- and query time that with high probability is polynomial in $\log n$ with no dependency on the arboricity. Thus, we improve the number of colours exponentially compared to the state-of-the art for implicit colouring, namely from $O(2^\alpha)$ colours, and we approach the theoretical lower bound of $\Omega(\alpha)$ for this arboricity-parameterised approach. Simultaneously, our randomised algorithm improves the update- and query time to run in time solely polynomial in $\log n$ with no dependency on $\alpha$. Our algorithms are fully adaptive to the current value of the dynamic arboricity at query or update time.