Scholarlink

To exploit both deep representation and kernel interpretability, we couple a deep neural network (DNN) classifier with KNMF clustering for single-cell RNA-seq data. A hierarchical representative-set strategy reduces input size while preserving rare cell types. Optimal representative set selection is realized through KL divergence minimization based on stratifying sampling. KNMF partitions the representative set and the DNN refines subtype discrimination. Benchmarked on multiple real-world datasets, the pipeline attains outstanding performance, demonstrating complementary strengths of kernel learning and deep learning.

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

Deep Neural Network with Kernel Nonnegative Matrix Factorization for Single Cell Clustering

  • 章节

  • 发表日期：

    2026年

查看全文
  • Hao Jiang,
  • Wai-Ki Ching

摘要

To exploit both deep representation and kernel interpretability, we couple a deep neural network (DNN) classifier with KNMF clustering for single-cell RNA-seq data. A hierarchical representative-set strategy reduces input size while preserving rare cell types. Optimal representative set selection is realized through KL divergence minimization based on stratifying sampling. KNMF partitions the representative set and the DNN refines subtype discrimination. Benchmarked on multiple real-world datasets, the pipeline attains outstanding performance, demonstrating complementary strengths of kernel learning and deep learning.