An Adaptive Alternating-direction-method-based Nonnegative Latent Factor Model
This work addresses scalability challenges in representation learning for industrial applications with high-dimensional incomplete data, but it is incremental as it builds on existing methods by adding hyper-parameter adaptation.
The paper tackled the issue of hyper-parameter tuning in alternating-direction-method-based nonnegative latent factor models for high-dimensional incomplete matrices, proposing an adaptive model (A2NLF) that uses particle swarm optimization and achieves improved computational and storage efficiency while maintaining competitive estimation accuracy.
An alternating-direction-method-based nonnegative latent factor model can perform efficient representation learning to a high-dimensional and incomplete (HDI) matrix. However, it introduces multiple hyper-parameters into the learning process, which should be chosen with care to enable its superior performance. Its hyper-parameter adaptation is desired for further enhancing its scalability. Targeting at this issue, this paper proposes an Adaptive Alternating-direction-method-based Nonnegative Latent Factor (A2NLF) model, whose hyper-parameter adaptation is implemented following the principle of particle swarm optimization. Empirical studies on nonnegative HDI matrices generated by industrial applications indicate that A2NLF outperforms several state-of-the-art models in terms of computational and storage efficiency, as well as maintains highly competitive estimation accuracy for an HDI matrix's missing data.