• Publications
  • Influence
Semi-Supervised Nonnegative Matrix Factorization
TLDR
This work presents semi-supervised NMF (SSNMF), where they jointly incorporate the data matrix and the (partial) class label matrix into NMF, and develops multiplicative updates for SSNMF to minimize a sum of weighted residuals.
Weighted Nonnegative Matrix Co-Tri-Factorization for Collaborative Prediction
TLDR
Weighted Nonnegative Matrix Co-Tri-Factorization (WNMCTF) is developed where the authors jointly minimize weighted residuals, each of which involves a nonnegative 3-factor decomposition of target or side information matrix.
Orthogonal Nonnegative Matrix Factorization: Multiplicative Updates on Stiefel Manifolds
TLDR
This paper presents an algorithm for orthogonal nonnegative matrix factorization, where an orthogonality constraint is imposed on the nonnegative decomposition of a term-document matrix, developed directly from true gradient on Stiefel manifold.
Nonnegative matrix partial co-factorization for drum source separation
TLDR
This paper presents nonnegative matrix partial co-factorization (NMPCF) where the target matrix and drum-only-matrix are simultaneously decomposed, sharing some factor matrix partially, to force some portion of basis vectors to be associated with drums only.
Nonnegative Matrix Factorization with Orthogonality Constraints
TLDR
An algorithm for orthogonal nonnegative matrix factorization, where an orthogonality constraint is imposed on the nonnegative decomposition of a term-document matrix, which can be clearly interpreted for the clustering problems.
Nonnegative Matrix Partial Co-Factorization for Spectral and Temporal Drum Source Separation
TLDR
Nonnegative matrix partial co-factorization (NMPCF) is applied to several spectrogram matrices in which column-blocks of mixture spectrograms and the drum-only matrix are jointly decomposed, sharing a factor matrix partially, in order to determine common basis vectors that capture spectral and temporal characteristics of drum sources.
Efficient learning of non-autoregressive graph variational autoencoders for molecular graph generation
TLDR
This paper presents an improved learning method involving a graph variational autoencoder for efficient molecular graph generation in a non-autoregressive manner and introduces three additional learning objectives: approximate graph matching, reinforcement learning, and auxiliary property prediction.
Probabilistic matrix tri-factorization
TLDR
An EM algorithm is developed to learn the PMTF model, showing its equivalence to multiplicative updates derived by an algebraic approach, and the useful behavior of PMTF is demonstrated in a task of document clustering.
Deep-learning-based inverse design model for intelligent discovery of organic molecules
TLDR
An inverse design model based on a deep encoder-decoder architecture for targeted molecular design based on neural machine language translation is developed, which successfully learned design rules from the given databases and generated promising light-absorbing molecules and host materials for a phosphorescent organic light-emitting diode by creating new ligands and combinatorial rules.
...
1
2
3
...