Research map / Probabilistic and scientific modeling / Treatment effects
Adapting Neural Networks for the Estimation of Treatment Effects
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add a propensity-score head to the TARNET-style model (ablation) | TARNET · ACIC 2018 | MAE: 1.45 -> 0.55 | Helped |
| Add targeted regularization (ablation) | Dragonnet · ACIC 2018 | MAE: 0.55 -> 0.35 | Helped |
| Add targeted regularization (ablation) | TARNET · ACIC 2018 | MAE: 1.45 -> 1.40 | Helped |
| Use an end-to-end treatment-and-outcome representation rather than staged treatment training with a frozen representation | Dragonnet versus NEDnet · ACIC 2018 | MAE: 1.49 -> 0.55 | Helped |
| Use targeted regularization rather than the unregularized model (ablation) | TARNET · IHDP Δ_in and Δ_all | MAE: 0.16 -> 0.15; 0.13 -> 0.12 | Mixed |
| Use targeted regularization rather than the unregularized model (ablation) | TARNET · IHDP Δ_out | MAE: 0.21 -> 0.20 | Helped |
| Use targeted regularization rather than the unregularized model (ablation) | Dragonnet · IHDP Δ_out | MAE: 0.21 -> 0.20 | Helped |
| Use targeted regularization rather than the unregularized model (ablation) | Dragonnet · IHDP Δ_in and Δ_all | MAE: 0.14 -> 0.14; 0.12 -> 0.11 | No clear effect |
| Use TMLE instead of the Q-only estimator | Dragonnet · ACIC 2018 | MAE: 0.55 -> 1.97 | Hurt |
| Add targeted regularization (ablation) | Dragonnet · ACIC 2018, baseline error Δ_ATE >1 | MAE: 10.64 -> 6.50 | Helped |
| Use TMLE instead of the Q-only estimator | TARNET · ACIC 2018, baseline error Δ_ATE >1 | MAE: 28.52 -> 176.14 | Hurt |
| Tighten propensity-score truncation from [0.01, 0.99] to [0.1, 0.9] (ablation) | TARNET with Q estimator · ACIC 2018 | MAE: 1.45 -> 0.44 | Helped |
Closest papers by meaning
- Adjustment for Confounding using Pre-Trained Representations (ICML 2025 · similarity 0.87)
- Learning Representations of Instruments for Partial Identification of Treatment Effects (ICML 2025 · similarity 0.86)
- PairNet: Training with Observed Pairs to Estimate Individual Treatment Effect (ICML 2024 · similarity 0.86)
- Doubly robust identification of treatment effects from multiple environments (ICLR 2025 · similarity 0.86)
- Treatment Effect Estimation for Optimal Decision-Making (NeurIPS 2025 · similarity 0.85)
- NESTER: An Adaptive Neurosymbolic Method for Causal Effect Estimation (AAAI 2024 · similarity 0.85)
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.