# New Amortized Cell-Probe Lower Bounds for Dynamic Problems

## TL;DR

This paper introduces a general framework for deriving amortized cell-probe lower bounds for dynamic problems, providing the strongest known unconditional bounds for specific problems like polynomial evaluation and matrix-vector multiplication.

## Contribution

It extends previous worst-case bounds to amortized bounds using a new framework, achieving the strongest unconditional lower bounds for dynamic data structures.

## Key findings

- Dynamic polynomial evaluation requires either high update or query time.
- Dynamic online matrix-vector multiplication requires either high update or query time.
- The bounds match the best known unconditional lower bounds for these problems.

## Abstract

We build upon the recent papers by Weinstein and Yu (FOCS'16), Larsen (FOCS'12), and Clifford et al. (FOCS'15) to present a general framework that gives amortized lower bounds on the update and query times of dynamic data structures. Using our framework, we present two concrete results. (1) For the dynamic polynomial evaluation problem, where the polynomial is defined over a finite field of size $n^{1+\Omega(1)}$ and has degree $n$, any dynamic data structure must either have an amortized update time of $\Omega((\lg n/\lg \lg n)^2)$ or an amortized query time of $\Omega((\lg n/\lg \lg n)^2)$. (2) For the dynamic online matrix vector multiplication problem, where we get an $n \times n$ matrix whose entires are drawn from a finite field of size $n^{\Theta(1)}$, any dynamic data structure must either have an amortized update time of $\Omega((\lg n/\lg \lg n)^2)$ or an amortized query time of $\Omega(n \cdot (\lg n/\lg \lg n)^2)$. For these two problems, the previous works by Larsen (FOCS'12) and Clifford et al. (FOCS'15) gave the same lower bounds, but only for worst case update and query times. Our bounds match the highest unconditional lower bounds known till date for any dynamic problem in the cell-probe model.