AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Path-Normalized Optimization of Recurrent Neural Networks with ReLU Activations

Behnam Neyshabur, Yuhuai Wu, Ruslan Salakhutdinov, Nati Srebro

NeurIPS 2016 · Architectures and efficiency › RNNs and normalization · approach: RNN training dynamics and gradients · 16 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train with Path-SGD instead of SGDsingle-layer RNN, 200 hidden units · word-level PTBtraining and test errorImproved
drop second term kappa(2) from Path-SGD update (ablation)Path-SGD · word-level PTBtraining and test errorNo clear effect
fewer hidden units H and longer length TPath-SGD second term · PTBnorm ratio of kappa(2) to kappa(1)Improved
Path-SGD with identity-initialized RNN instead of SGDIRNN · Adding problem T=400test MSE: 16.7 -> 0Improved
Path-SGD with identity-initialized RNN instead of SGDIRNN · Adding problem T=750test MSE: 16.7 -> 0Improved
increase sequence length to 1000RNN-Path · Adding problemlearning successHurt
Path-SGD instead of SGD for identity-initialized RNNIRNN · sequential MNISTtest error: 7.1 -> 3.1Improved
Path-SGD instead of plain training of ReLU RNNRNN-ReLU · PTB character-leveltest BPC: 1.55 -> 1.47Improved
Path-SGD instead of plain training of ReLU RNNRNN-ReLU · text8test BPC: 1.65 -> 1.58Improved
Path-SGD ReLU RNN versus tanh RNNRNN-tanh · text8test BPC: 1.70 -> 1.58Improved
Path-SGD RNN versus TRec beta=500RNN-Path · PTB character-leveltest BPC: 1.48 -> 1.47Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy