Optimal Training-Time Scaling in Gradual Adaptation

arXiv:2608.04927v1 Announce Type: new Abstract: In gradual adaptation, how should the training time on each task change as the number of intermediate tasks increases? We study this question for overparameterized linear regression tasks that change smoothly and share a zero-loss solution. With $N$ tasks and training time $s_N$ on each, the final learning progress converges to a continuum curve when $Ns_N\to\tau$. The limiting progress is $\Theta(\tau)$ for small $\tau$ and $\Theta(\tau^{-1})$ for large $\tau$, so both very short and very long training produce little progress. It follows that optimal per-task training times scale as $s_N^\star=\Theta(N^{-1})$, equivalently $Ns_N^\star=\Theta(1)$. Experiments on gradually rotated MNIST and a natural Yearbook time shift are consistent with less per-task training as the path is divided more finely.

Abstract:In gradual adaptation, how should the training time on each task change as the number of intermediate tasks increases? We study this question for overparameterized linear regression tasks that change smoothly and share a zero-loss solution. With $N$ tasks and training time $s_N$ on each, the final learning progress converges to a continuum curve when $Ns_N\to\tau$. The limiting progress is $\Theta(\tau)$ for small $\tau$ and $\Theta(\tau^{-1})$ for large $\tau$, so both very short and very long training produce little progress. It follows that optimal per-task training times scale as $s_N^\star=\Theta(N^{-1})$, equivalently $Ns_N^\star=\Theta(1)$. Experiments on gradually rotated MNIST and a natural Yearbook time shift are consistent with less per-task training as the path is divided more finely.

Machine Learning (cs.LG)

arXiv:2608.04927 [cs.LG]

(or arXiv:2608.04927v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2608.04927 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

analiza aiqbench

Optimal Training-Time Scaling in Gradual Adaptation

Subiecte conexe