On Bock's Conjecture Regarding the Adam Optimizer

11/16/2021
by   Mohamed Akrout, et al.
11

In 2014, Kingma and Ba published their Adam optimizer algorithm, together with a mathematical argument that was meant to help justify it. In 2018, Bock and colleagues reported that a key piece was missing from that argument - an unproven lemma which we will call Bock's conjecture. Here we show that this conjecture is false, but a modified version of it does hold, and fills the gap in Bock's proof of convergence for Adam.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
08/17/2021

A Note on the Permuted Puzzles Toy Conjecture

In this note, we show that a "Toy Conjecture" made by (Boyle, Ishai, Pas...
research
07/20/2019

A short proof of a conjecture by Ben-Akiva and Lerman about the nested logit model

We provide a short proof of a result by Cardell (1997) on a conjecture o...
research
02/18/2022

On The "Majority is Least Stable" Conjecture

We show that the "majority is least stable" conjecture is true for n=1 a...
research
06/17/2022

A remark on Kashin's discrepancy argument and partial coloring in the Komlós conjecture

In this expository note, we discuss an early partial coloring result of ...
research
07/20/2019

On a conjecture by Ben-Akiva and Lerman about the nested logit model

We prove a conjecture of Ben-Akiva and Lerman (1985) regarding the rando...
research
04/07/2019

On the convergence proof of AMSGrad and a new version

The adaptive moment estimation algorithm Adam (Kingma and Ba, ICLR 2015)...
research
10/31/2019

Computational pseudorandomness, the wormhole growth paradox, and constraints on the AdS/CFT duality

A fundamental issue in the AdS/CFT correspondence is the wormhole growth...

Please sign up or login with your details

Forgot password? Click here to reset