Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Fast Instrument Learning with Faster Rates
Ziyu Wang, Yuhao Zhou, Jun Zhu
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace RBF-kernel first stage with learned DNN instrument features (ablation) | Kernelized IV · f0~GP, D=100, n1=n2=2500 | Test MSE: .176 -> .035 | Improved |
| replace RBF-kernel first stage with learned DNN instrument features (ablation) | Kernelized IV · f0~GP, D=100, n1=n2=500 | 90% credible ball coverage: .187 -> .923 | Improved |
| replace RBF first stage with learned instruments (ablation) | Kernelized IV · f0~GP, D=100, n1=n2=5000 | 90% credible ball radius: .156 -> .049 | Improved |
| increase sample size n1=n2 from 500 to 5000 | Proposed · f0~GP, D=100 | Test MSE: .097 -> .024 | Improved |
| increase sample size n1=n2 from 500 to 5000 | Proposed · f0~GP, D=100 | 90% credible ball coverage: .923 -> .920 | No clear effect |
| replace adaptive AGMM-NN/Tree with proposed black-box + kernel method | Proposed · predictive, N=2500, D=100 | runtime (s): 303 -> 25.9 | Improved |
| use AGMM-RBF instead of proposed | AGMM · predictive, N=2500, D=100 | runtime (s): 25.9 -> 6.7 | Mixed |
| increase dimension D (instruments) for AGMM-Tree and AGMM-NN | AGMM-Tree/NN · predictive | test MSE | Hurt |
| trained DNN first stage vs RBF kernel or random-init DNN (ablation) | Instrument learning · f0~GP, D in {2,40,100} | counterfactual MSE | Improved |
| use proposed method vs AGMM-NN and prior MNIST baseline | Proposed · MNIST/CIFAR image instruments | test MSE | Improved |
Closest papers by meaning
- Stochastic Optimization Algorithms for Instrumental Variable Regression with Streaming Data (NeurIPS 2024 · similarity 0.87)
- Optimality and Adaptivity of Deep Neural Features for Instrumental Variable Regression (ICLR 2025 · similarity 0.87)
- Nonparametric Instrumental Variable Regression through Stochastic Approximate Gradients (NeurIPS 2024 · similarity 0.85)
- Demystifying Spectral Feature Learning for Instrumental Variable Regression (NeurIPS 2025 · similarity 0.85)
- Geometry-Aware Instrumental Variable Regression (ICML 2024 · similarity 0.84)
- Learning Decision Policies with Instrumental Variables through Double Machine Learning (ICML 2024 · similarity 0.83)
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.
