Scholarlink

<p>Transfer learning addresses the challenge of transferring knowledge from one domain to another. Traditional transfer learning focuses on adapting models trained on a source domain (with many observations) to improve performance on a target domain (with few observations). In this work, we consider the case of a model shift and focus on transfer learning applied to a causal forest, namely HTERF. This causal forest aims to estimate the Conditional Average Treatment Effect (CATE). The approach considered is the offset method presented by Wang (<CitationRef CitationID="CR17">2016</CitationRef>), adapted to a causal context. This method relies on the use of intermediate models to estimate the offset between source and target distributions. We establish an <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(L^1\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>L</mi> <mn>1</mn> </msup> </math></EquationSource> </InlineEquation>-consistency result of our algorithm and derive a bound on the CATE estimation error of HTERF in the target domain, depending on the error of the intermediate models. Simulation studies demonstrate the good performance of this approach in different settings.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Transfer learning for causal forests

  • 期刊论文

  • 发表日期：

    2026年6月26日

查看全文
  • Bérénice-Alexia Jocteur,
  • Véronique Maume-Deschamps,
  • Pierre Ribereau

摘要

Transfer learning addresses the challenge of transferring knowledge from one domain to another. Traditional transfer learning focuses on adapting models trained on a source domain (with many observations) to improve performance on a target domain (with few observations). In this work, we consider the case of a model shift and focus on transfer learning applied to a causal forest, namely HTERF. This causal forest aims to estimate the Conditional Average Treatment Effect (CATE). The approach considered is the offset method presented by Wang (2016), adapted to a causal context. This method relies on the use of intermediate models to estimate the offset between source and target distributions. We establish an \(L^1\) L 1 -consistency result of our algorithm and derive a bound on the CATE estimation error of HTERF in the target domain, depending on the error of the intermediate models. Simulation studies demonstrate the good performance of this approach in different settings.