Scholarlink

<p>We introduce a principled approach to synthetic data generation, leveraging a fully connected neural network ((multilayer perceptron) with piecewise linear activation functions to address a unary classification task. We show that the network’s output serves as an adaptive histogram-based estimation of the probability density function over a compact domain. The proposed method constructs synthetic tabular datasets by filtering uniformly distributed random vectors based on the inferred empirical density. The effectiveness of this approach is illustrated through a series of model experiments. Bibliography: 10 titles. Illustrations: 6 figures.</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

CONSISTENT METHOD FOR SYNTHETIC TABULAR DATA OBTAINING USING A MULTILAYER PERCEPTRON

  • 期刊论文

  • 发表日期：

    2025年11月1日

查看全文
  • Andrey Perminov,
  • Andrey Kovalenko,
  • Denis Turdakov

摘要

We introduce a principled approach to synthetic data generation, leveraging a fully connected neural network ((multilayer perceptron) with piecewise linear activation functions to address a unary classification task. We show that the network’s output serves as an adaptive histogram-based estimation of the probability density function over a compact domain. The proposed method constructs synthetic tabular datasets by filtering uniformly distributed random vectors based on the inferred empirical density. The effectiveness of this approach is illustrated through a series of model experiments. Bibliography: 10 titles. Illustrations: 6 figures.