GO Hessian for Expectation-Based Objectives

Authors: Yulai Cong, Miaoyun Zhao, Jianqiao Li, Junya Chen, Lawrence Carin12060-12068

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Leveraging the GO Hessian, we develop a new second-order method for Eqγ (y)[f(y)], with challenging experiments conducted to verify its effectiveness and efficiency. The proposed techniques are verified with challenging experiments where non-rep gamma RVs are of interest.
Researcher Affiliation Academia Yulai Cong, Miaoyun Zhao,* Jianqiao Li, Junya Chen, Lawrence Carin Department of Electrical and Computer Engineering, Duke University
Pseudocode Yes Algorithm 1 SCR-GO for minγ Eq(x)qγ(y|x)[f(x, y)]
Open Source Code Yes Code will be available at github.com/YulaiCong/GOHessian.
Open Datasets Yes Variational Encoders for PFA and PGBN... Figures 4(b)-4(c) show the training objectives versus the number of oracle calls and processed observations... PFA on MNIST.
Dataset Splits No The paper describes using training datasets and showing training curves but does not explicitly provide details on train/validation/test splits (e.g., percentages, sample counts, or specific predefined splits).
Hardware Specification Yes The Titan Xp GPU used was donated by the NVIDIA Corporation.
Software Dependencies No The paper mentions software like PyTorch and TensorFlow but does not specify their version numbers or any other software dependencies with specific versions.
Experiment Setup Yes For both SGD and Adam, learning rates from {0.001, 0.005, 0.01, 0.05, 0.1, 0.5, 1} are tested with the best-tuned results shown. Other settings are given in Appendix J.