{"id":"arxiv/2606.08188","url":"https://arxiv.org/abs/2606.08188","tags":["math.OC","cs.LG"],"score":0,"title":"Latent Structural Categorical Matrix Completion with Application to Quasispecies Analysis","domain":"math.OC","source":"arxiv","authors":["Qian Zhang","Meixia Lin"],"reviews":0,"subject":"math.OC","summary":"Matrix completion has been extensively studied for real-valued data, but existing methods are often limited in handling categorical variables. We propose LCMC, a double-loop optimization framework for categorical matrix completion via latent factorization based on a binary tensor representation. In this setting, each categorical entry is encoded as a one-hot vector along a third tensor mode, thereby preserving its discrete, non-ordinal nature. The outer loop adaptively estimates the latent dimension by iteratively updating it with feedback from the inner loop, while the inner loop…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-06-06","reactions":{}}