AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

HOPE: High-Order Polynomial Expansion of Black-Box Neural Networks

Tingxiong Xiao, Weihang Zhang, Yuxiao Cheng, Jinli Suo

TPAMI 2024 · Learning theory and algorithms › Prediction and decision losses · approach: Provable post-hoc explanations

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
use HOPE instead of Autograd for 10th-order derivatives10-layer MLP, 1-D input · 1-D input MLP, order 10time (s): 5.4099 -> 0.1942Improved
use HOPE instead of Autograd for 5th-order derivatives10-layer MLP, 3-D input · 3-D input MLP, order 5time (s): 6.9751 -> 0.0711Improved
use HOPE instead of Autograd for 6th-order unmixed derivativesMNIST 01 classifier · MNIST 01 classifier, order 6time (s): 1129.2 -> 0.0479Improved
use HOPE where Autograd runs out of memoryMNIST 01 classifier · MNIST 01 classifier, order 7-10memory: OOM / 76+ -> 0.1000 / 0.30Improved
replace network with 3rd-order Taylor polynomiallevel controller · batch 4096 inferenceinference time (ms): 121.98 -> 0.0324Improved
replace network with 3rd-order Taylor polynomiallevel controller · model sizebytes: 44,671 -> 160Improved
replace perturbation analysis with HOPE heat mapMNIST classifier · heat map generationtime (s): 0.342 -> 0.002Improved
reduce weight scale w0 from 1.0 to 0.01 (ablation)MLP · MLP width 512 depth 510th-order derivative ratio: 1.39e-01 -> 1.31e-19Improved
increase weight scale w0 from 1.0 to 100 (ablation)MLP · MLP width 512 depth 510th-order derivative ratio: 1.39e-01 -> 1.23e+21Hurt
use HOPE for local approximation of ReLU and max-pool networksHOPE · networks with ReLU or max poolingderivative order obtainedNo clear effect

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