MSAFIS: an evolving fuzzy inference system

Abstract

In this paper, the problem of learning in big data is considered. To solve this problem, a new algorithm is proposed as the combination of two important evolving and stable intelligent algorithms: the sequential adaptive fuzzy inference system (SAFIS), uniform stable backpropagation algorithm (SBP). The modified sequential adaptive fuzzy inference system (MSAFIS) is the SAFIS with the difference that the SBP is used instead of the Kalman filter for the updating of parameters. The SBP has the advantage that is obtains a better learning than the Kalman filter in big data. The effectiveness of the studied methods is verified by two experiments.

Learning evolving T–S fuzzy systems with both local and global accuracy – A local online optimization approach

“…[11] proposes a robust 70 online adaptive method called SOFMLS to identify Mamdani fuzzy systems by adding new fuzzy rules using a distance based approach and pruning fuzzy rules according to the density of the clusters. Similar researches could also be found in [12,13,[25][26][27][28][29][30][31][32][33].…”

“…Let condition 4.3 be satisfied, and the l 1 -th and the l 2 -th fuzzy rule be merged to l 0 -th (l 0 = l 1 numerically) rule at time t + 1. Then ψ i t+1 which minimizes (6) could be updated by (27) to (30),…”

A Self-Evolving Fuzzy System Which Learns Dynamic Threshold Parameter by Itself

“…The numerical results are evaluated by the rooted mean square errors (RMSEs(31)) (e.g. [3], [12], [15], [19], [20], [24]), and the non-dimensional error indexes (NDEIs (32)) (e.g. [3], [5], [15], [22], [23], [26]).…”

“…In this paper, the MATLAB built-in function er f (x), which is a rational function approximation shown in [40], is used for approximating the error function. Similar to the first term of (12), the second term of (12) could be computed by (18), (19) gives how to compute the third term in (12),…”

Online Interval Type-2 Fuzzy Extreme Learning Machine applied to 3D path following for Remotely Operated Underwater Vehicles

“…To compare the performance of the proposed OIT2-FELM with respect to other existing fuzzy methodologies, an OIT2-FELM using three different output layers is suggested, namely: an output layer with an SC algorithm, with a NT approach and with an EKM algorithm respectively. For the nonlinear system identification, in Table 2, a comparison of the average performance of ten experiments between the proposed OIT2-FELM and other learning fuzzy methodologies such as ANFIS model [58], Sequential-Adaptive-Fuzzy-Inference-System (SAFIS) [59], and the classical version of sequential Online FELM (OS-FELM) [44]. For the implementation of OIT2-FELM, two different types of MFs are implemented, i.e.…”