We study the generalization capability of a modular neural network`the mixture of experts' that learn from examples generated by another network with the same architecture. When the number of examples is smaller than a critical value, the network shows a symmetric phase where the role of the experts is not specialized. Upon crossing the critical point, the system undergoes a continuous phase tr...

Several experiments were conducted in order to investigate the usefulness of mixture of experts (ME) approach to an online internet system assisting in real estate appraisal. All experiments were performed using 28 realworld datasets composed of data taken from a cadastral system and GIS data derived from a cadastral map. The analysis of the results was performed using recently proposed statist...

We introduce and analyze LocBoost, a new boosting algorithm, which leads to the incremental construction of a mixture of experts type architecture. We provide upper bounds on the expected loss of such models in terms of the smoothness properties of the gating functions appearing in the mixture of experts model. Furthermore, an incremental algorithm is proposed for the construction of the classi...

The long-tailed video recognition problem is especially challenging, as videos tend to be long and untrimmed, each may contain multiple classes, causing frame-level class imbalance. previous method tackles the only through sampling for re-balance without distinguishing feature representation between head tail classes. To improve of we modulate features with an auxiliary distillation loss reduce...

Accurate estimation of multiple quality variables is critical for building industrial soft sensor models, which have long been confronted with data efficiency and negative transfer issues. Methods sharing backbone parameters among tasks address the issue; however, they still fail to mitigate problem. To this issue, a balanced Mixture-of-Experts (BMoE) proposed in work, consists multi-gate mixtu...