AI Paper Atlas

Research map / Optimization / Optimizers

Analytically Tractable Hidden-States Inference in Bayesian Neural Networks

Luong-Ha Nguyen, James-A. Goulet

JMLR 2022 · Optimization › Optimizers · approach: Natural gradient and Fisher approximations

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
apply TAGI hidden-state inference on the input image toward a chosen target label2 conv. CNN on MNIST · MNISTerror rate (%): 0.64 -> 99.8Improved
apply TAGI hidden-state inference as a non-targeted attack2 conv. CNN on MNIST · MNISTerror rate (%): 0.64 -> 99.9Improved
apply TAGI hidden-state inference as a targeted attack3 conv. CNN on Cifar10 · Cifar10error rate (%): 22.0 -> 99.6Improved
apply TAGI hidden-state inference as a non-targeted attack3 conv. CNN on Cifar10 · Cifar10error rate (%): 22.0 -> 99.9Improved
add positive derivative constraint alpha=+1 (seek minimum), start 0.25 (ablation)TAGI inference-based optimization on 1D toy function · y=x^3-3xinferred optimal location: 0.965 -> -0.993Improved
add negative derivative constraint alpha=-1 (seek maximum), start -0.25 (ablation)TAGI inference-based optimization on 1D toy function · y=x^3-3xinferred optimal location: -0.992 -> 0.965Improved
start optimization at 1.9 instead of near the optimum (ablation)TAGI inference-based optimization · y=x^3-3xreaching local maximum at x=-1Hurt
use TAGI inference of actions instead of PPO gradient ascentcontinuous-action RL · Half-cheetah (Mujoco)average reward: 1649 -> 1934Improved
use TAGI inference of actions instead of PPO gradient ascentcontinuous-action RL · Inverted pendulum (Mujoco)average reward: 887 -> 983Improved

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