Mixed counting models that use the negative binomial distribution as the prior can well model over-dispersed and hierarchically dependent random variables; thus they have attracted much attention in mining dispersed document topics. However, the existing parameter inference method like Monte Carlo sampling is quite time-consuming. In this paper, we propose two efficient neural mixed counting models, i.e., the Negative Binomial-Neural Topic Model (NB-NTM) and the Gamma Negative Binomial-Neural Topic Model (GNB-NTM) for dispersed topic discovery. Neural variational inference algorithms are developed to infer model parameters by using the reparameterization of Gamma distribution and the Gaussian approximation of Poisson distribution. Experiments on real-world datasets indicate that our models outperform state-of-theart baseline models in terms of perplexity and topic coherence. The results also validate that both NB-NTM and GNB-NTM can produce explainable intermediate variables by generating dispersed proportions of document topics. * The first two authors contributed equally to this work which was finished when Jiemin Wu was an undergraduate student of his final year.† The corresponding author.I'd prefer e-mailed responses as I don't get to read this newsgroup often.

Topic modeling has been widely used for discovering the latent semantic structure of documents, but most existing methods learn topics with a flat structure. Although probabilistic models can generate topic hierarchies by introducing nonparametric priors like Chinese restaurant process, such methods have data scalability issues. In this study, we develop a tree-structured topic model by leveraging nonparametric neural variational inference. Particularly, the latent components of the stickbreaking process are first learned for each document, then the affiliations of latent components are modeled by the dependency matrices between network layers. Utilizing this network structure, we can efficiently extract a tree-structured topic hierarchy with reasonable structure, low redundancy, and adaptable widths. Experiments on real-world datasets validate the effectiveness of our method.

Topic models have been widely used for learning the latent explainable representation of documents, but most of the existing approaches discover topics in a flat structure. In this study, we propose an effective hierarchical neural topic model with strong interpretability. Unlike the previous neural topic models, we explicitly model the dependency between layers of a network, and then combine latent variables of different layers to reconstruct documents. Utilizing this network structure, our model can extract a tree-shaped topic hierarchy with low redundancy and good explainability by exploiting dependency matrices. Furthermore, we introduce manifold regularization into the proposed method to improve the robustness of topic modeling. Experiments on real-world datasets validate that our model outperforms other topic models in several widely used metrics with much fewer computation costs.

Background. To analyze the effects of the implementation of emergency surgical patterns in patients with rhegmatogenous retinal detachment (RRD) and provide evidence for promoting emergency surgical patterns for RRD. Methods. We reviewed the electronic medical records of 346 patients (348 eyes) who underwent surgical repair of RRD at the Zhongshan Ophthalmic Center in Southern China. A total of 140 patients (140 eyes) in the routine inpatient surgery group were collected at the fundus disease department between January 2019 and December 2019, and 206 patients (208 eyes) in the emergency surgery group were collected at the ophthalmic emergency department between January 2021 and December 2021. Demographics, best-corrected visual acuity (BCVA) expressed as the logarithm of the minimum angle of resolution (logMAR), the status of the macula before surgery, time to presentation, treatment interval, and postoperative BCVA measured at least three months follow-up were compared. Results. The preoperative BCVA (logMAR) of the emergency surgery group and the inpatient surgery group were 1.0 (0.4–1.7) and 1.4 (0.7–1.7), respectively, with significant differences between groups ( P < 0.001 ). However, patients had a shorter time to presentation (7 days vs. 21 days, P < 0.001 ), shorter treatment interval (2 days vs. 12 days, P < 0.01 ), and significantly better postoperative BCVA (logMAR 0.5 vs. logMAR 1.0, P < 0.001 ) in the emergency surgery group than in the inpatient surgery group. There was no significant difference in primary anatomical success between the two groups ( P = 0.802 ). The median follow-up for the emergency surgery group and the inpatient surgery group were 6.08 months and 6.2 months, respectively, with no significant differences ( P > 0.05 ). Conclusions. Patients who underwent emergency surgical patterns of RRD had better visual outcomes after surgery than patients with routine inpatient surgery, which might be attributed to a shorter duration, shorter treatment interval, and the preoperative status of the macula in the emergency surgery pattern. Emergency surgical patterns for RRD should be considered to achieve better surgical outcomes in suitable patients.

Purpose To evaluate the safety and efficacy of pars plana vitrectomy (PPV) for recalcitrant vitreous seeding from retinoblastoma and concurrent retinal detachment. Methods In this retrospective interventional case series, intravitreal chemotherapy-assisted PPV was performed in 6 patients with persistent retinoblastoma vitreous seeds and concurrent retinal detachment. Globe salvage, retinal reattachment, and visual preservation at last follow-up visit were calculated. Results Six eyes, initially classified as group D (n=4) or E (n=2) with sphere and (or) cloud vitreous seeds, received standard systemic with intravitreal chemotherapy and subsequently developed refractory vitreous seeds with concurrent retinal detachment. PPV produced complete elimination of vitreous seeds in all 6 eyes and no vitreous seeding recurred over a mean follow-up of 29.6 months. Globe salvage and complete retina reattachment were attained in all 6 eyes (100%). Last follow-up vision acuity was assessable in 4 eyes with 80% (3/4) better than 0.4. Three eyes (50%) received repeated systemic or focal treatment because of the recurrent retinal tumor. There was no case of extraocular tumor extension or systemic metastasis. Conclusion Administration of intravitreal chemotherapy-assisted PPV appears to be a safe and effective alternation for the initial treatment of refractory vitreous seeds from retinoblastoma. It is also a promising therapeutic strategy to repair concurrent retinal detachment and preserve the remaining visual function.