SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2002.04839 · 2020

LaProp: Separating Momentum and Adaptivity in Adam

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 7 functions out of this paper's own repositories and ran 6 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
Z-T-WANG/LaProp-Optimizer canonical 6 of 7
FunctionStatusWhere it lives
beta_scheduler Ran Z-T-WANG/LaProp-Optimizer/rainbow/common/utils.py
code served (permissive licence) · get_code("11fc51b3483204eb")
create_log_dir Ran Z-T-WANG/LaProp-Optimizer/rainbow/common/utils.py
code served (permissive licence) · get_code("878f223c996c5522")
epsilon_scheduler Ran Z-T-WANG/LaProp-Optimizer/rainbow/common/utils.py
code served (permissive licence) · get_code("2e0c79ff61cfd9ae")
make_atari Ran Z-T-WANG/LaProp-Optimizer/rainbow/common/wrappers.py
code served (permissive licence) · get_code("6eb2fa89d6e4a36b")
wrap_deepmind Ran Z-T-WANG/LaProp-Optimizer/rainbow/common/wrappers.py
code served (permissive licence) · get_code("34023060eaec34a2")
wrap_pytorch Ran Z-T-WANG/LaProp-Optimizer/rainbow/common/wrappers.py
code served (permissive licence) · get_code("d818d11d4b5f6381")
DQN Not yet run Z-T-WANG/LaProp-Optimizer/rainbow/model.py
code served (permissive licence) · get_code("ed2dc93a9723ec69")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We identity a by-far-unrecognized problem of Adam-style optimizers which results from unnecessary coupling between momentum and adaptivity. The coupling leads to instability and divergence when the momentum and adaptivity parameters are mismatched. In this work, we propose a method, Laprop, which decouples momentum and adaptivity in the Adam-style methods. We show that the decoupling leads to greater flexibility in the hyperparameters and allows for a straightforward interpolation between the signed gradient methods and the adaptive gradient methods. We experimentally show that Laprop has consistently improved speed and stability over Adam on a variety of tasks. We also bound the regret of Laprop on a convex problem and show that our bound differs from that of Adam by a key factor, which demonstrates its advantage.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2002.04839")
get_code_for_paper("2002.04839")
have("2002.04839")

Connect an agent — have() is free.