April 6, 2021 08:52

Taiji Suzuki, the leader and the former visiting researcher of Deep Learning Theory Team wins the award entitled “ICLR 2021 Outstanding Paper Award” in the ICLR 2021, a major conference on machine learning.

Title: Optimal Rates for Averaged Stochastic Gradient Descent under Neural Tangent Kernel Regime
Authors:
Atsushi Nitanda(The University of Tokyo* / RIKEN AIP)
Taiji Suzuki(The University of Tokyo / RIKEN AIP)
*Oct 2, 2020(the deadline for the conference)
*His current institution: Kyushu Institute of Technology

For more details, please refer to the link below.

https://iclr-conf.medium.com/announcing-iclr-2021-outstanding-paper-awards-9ae0514734ab

ICLR2021 selected 8 papers as the award winners.
The number of posted papers was 2997. Among them, 860 papers were selected for the conference.

These charts show that “Correlations between the true function and the tangent kernels corresponding to the first and second layers, and its relation to learning speed”.

 

Related Link

ICLR 2021 Outstanding Paper Award

Related Laboratories

last updated on April 15, 2021 14:22Laboratory