Broad Learning System: An Effective and Efficient Incremental Learning System Without the Need for Deep Architecture

Abstract

Broad Learning System (BLS) that aims to offer an alternative way of learning in deep structure is proposed in this paper. Deep structure and learning suffer from a time-consuming training process because of a large number of connecting parameters in filters and layers. Moreover, it encounters a complete retraining process if the structure is not sufficient to model the system. The BLS is established in the form of a flat network, where the original inputs are transferred and placed as "mapped features" in feature nodes and the structure is expanded in wide sense in the "enhancement nodes." The incremental learning algorithms are developed for fast remodeling in broad expansion without a retraining process if the network deems to be expanded. Two incremental learning algorithms are given for both the increment of the feature nodes (or filters in deep structure) and the increment of the enhancement nodes. The designed model and algorithms are very versatile for selecting a model rapidly. In addition, another incremental learning is developed for a system that has been modeled encounters a new incoming input. Specifically, the system can be remodeled in an incremental way without the entire retraining from the beginning. Satisfactory result for model reduction using singular value decomposition is conducted to simplify the final structure. Compared with existing deep neural networks, experimental results on the Modified National Institute of Standards and Technology database and NYU NORB object recognition dataset benchmark data demonstrate the effectiveness of the proposed BLS.

Age of Information Aware Radio Resource Management in Vehicular Networks: A Proactive Deep Reinforcement Learning Perspective

“…Table II lists other parameter values. 6 The tradeoff between the performance improvement and the time spent for the offline training process with a more complex neural network architecture is still an open problem [47], [48]. During the implementation of first three baselines, after the centralized frequency band allocation at the RSU at the beginning of each scheduling slot j, the vTx of each VUE-pair k ∈ K transmits a maximum feasible number min{X j k , R j k,(max) } of packets [24].…”

DWnet: Deep-wide network for 3D action recognition

“…Dang and Wang et al [28] applied broad learning system to estimate cement compressive strength from microstructure images of cement. In [29], BLS was utilized for estimating sun visibility from given outdoor images, and then incremental broad learning system, the improvement of broad learning system proposed in [20], was applied to classify the sun visibility into more categories. Considering BLS is able to model the large-scale data, Shi and Wei et al [30] designed fisher broad learning system (FBLS), and adopted Local Log-Euclidean Multivariate Gaussian (L2EMG) [31] for student gesture recognition task.…”

Patch Learning

“…2) Use a single broader (wider) model. For example, when the performance of a simple multi-layer perceptron neural network is not enough, more nodes can be added to each hidden layer [8], or enhancement nodes can be added to convert it into a broad learning system [12], [13]. When the performance of a simple fuzzy system is not enough, more membership functions (MFs) or rules can be added to make it wider [14]; or, in other words, to sculpt the state space more finely [15].…”