Learn More
Unsupervised feature learning has emerged as a promising tool in learning representations from unlabeled data. However, it is still challenging to learn useful high-level features when the data contains a significant amount of irrelevant patterns. Although feature selection can be used for such complex data, it may fail when we have to build a learning(More)
While determining model complexity is an important problem in machine learning, many feature learning algorithms rely on cross-validation to choose an optimal number of features, which is usually challenging for online learning from a massive stream of data. In this paper, we propose an incremen-tal feature learning algorithm to determine the optimal model(More)
1 Discussion of the Update Rules We introduce the update rules that were considered in this work and discuss the robustness of each update rule in relation to its hyperparameters. 1.1 Update rule based on heuristics Roughly speaking, this update rule is based on the following idea: increase the number of feature increments when the performance improves(More)
For InAs/GaAs(001) quantum dot (QD) system, the wetting layer (WL) evolution and its temperature dependence were studied using reflectance difference spectroscopy and were analyzed with a rate equation model. WL thicknesses showed a monotonic increase at relatively low growth temperatures but showed an initial increase and then decrease at higher(More)
1. Semi-supervised PGBM There are many classification tasks where we are given a large number of unlabeled examples in addition to only a few labeled training examples. For such scenario , it is important to include unlabeled examples during the training to generalize well to the unseen data, and thus avoid overfitting. Larochelle and Ben-gio (2008)(More)
Analysis of the fictitious domain method with H 1-penalty for parabolic problem Abstract. We consider the fictitious domain method with H 1-penalty for a parabolic problem. First, the sharp error estimate for the H 1-penalty problem approximating the original problem is derived. Then, we present some regularity analysis and prior estimate to the H 1-penalty(More)
Analysis of the fictitious domain method with L 2-penalty for elliptic and parabolic problems Abstract. The fictitious domain method with L 2-penalty for elliptic and parabolic problems are considered, respectively. The regularity theorems and a priori estimates for L 2-penalty problems are given. We derive error estimates for penalization and finite(More)