A Deep Non-negative Matrix Factorization Algorithm for Acupuncture Omics Data Clustering.

Acupuncture omics data analysis enables a deep understanding of the physiological mechanisms and therapeutic effects of acupuncture. However, such omics data are often high-dimensional and require non-negativity constraints, posing significant challenges for analysis methods. To address these issues, we propose a deep non-negative matrix factorization algorithm for acupuncture omics data clustering (DNMOC). First, the omics data is transformed into low-dimensional matrices by non-negative matrix factorization (NMF). Second, these low-dimensional matrices elements are non-negatively restricted using an activation function and updated by stochastic gradient descent method. Third, the gradient values corresponding to element updates are transformed into biases and weights, which are combined with the nonlinear function to construct the DNMOC network. Fourth, the DNMOC network is used to realize the learning of the low-dimensional matrices. Finally, extensive experiments on six acupuncture omics datasets demonstrate that our algorithm outperforms contemporary methods.