Narrow Results

A multi-feature broad learning system (MFBLS) is proposed to improve the image classification performance of broad learning system (BLS) and its variants. The model is characterized by two major characteristics: multi-feature extraction method and parallel structure. Multi-feature extraction method is utilized to improve the feature-learning ability of BLS. The method extracts four features of the input image, namely convolutional feature, K-means feature, HOG feature and color feature. Besides, a parallel architecture that is suitable for multi-feature extraction is proposed for MFBLS. There are four feature blocks and one fusion block in this structure. The extracted features are used directly as the feature nodes in the feature block. In addition, a “stacking with ridge regression” strategy is applied to the fusion block to get the final output of MFBLS. Experimental results show that MFBLS achieves the accuracies of 92.25%, 81.03%, and 54.66% on SVHN, CIFAR-10, and CIFAR-100, respectively, which outperforms BLS and its variants. Besides, it is even superior to the deep network, convolutional deep belief network, in both accuracy and training time on CIFAR-10. Code for the paper is available at https://github.com/threedteam/mfbls.

Parallel structure is a way to factor out common constituents in the expressions, which makes an effect of simplification of expressions. The complexity can be greatly reduced at the phase of sentence parsing by identifying such boundaries of parallel structure.

In this paper, we propose a probabilistic model to identify parallel cores (corresponding constituents) as well as boundaries of parallel noun phrases conjoined by "wa/gwa" (conjunctive particle in Korean). It is based on the idea of swapping constituents, utilizing symmetry (two or more identical constituents are repeated) and reversibility (the order of constituents is changeable) in parallel structure. The probabilities are calculated from (unlabelled) corpus with parallel structures, which is an advantage over the approaches trained with labeled corpus. Our model, moreover, is not dependent on languages.

It is also shown that the semantic features of the modifiers around parallel noun phrase and the patterns among words can be utilized further to correct the boundaries identified by the swapping model.

Experiment shows that our probabilistic swapping model performs much better than symmetry-based model and machine learning based approaches.

The subject of investigations are the almost hypercomplex manifolds with Hermitian and anti-Hermitian (Norden) metrics. A linear connection D is introduced such that the structure of these manifolds is parallel with respect to D and its torsion is totally skew-symmetric. The class of the nearly Kähler manifolds with respect to the first almost complex structure is of special interest. It is proved that D has a D-parallel torsion and is weak if it is not flat. Some curvature properties of these manifolds are studied.

A natural connection with totally skew-symmetric torsion on almost contact manifolds with B-metric is constructed. The class of these manifolds, where the considered connection exists, is determined. Some curvature properties for this connection, when the corresponding curvature tensor has the properties of the curvature tensor for the Levi-Civita connection and the torsion tensor is parallel, are obtained.