Name
Abstract | ||
|---|---|---|
Although Deep Belief Network (DBN) has been applied to a wide range of practical scenarios, i.e. image classification, signal recognition, remaining useful life estimation, on account of its powerful high classification accuracy, but it has impossible interpretation of functionality (it is desirable to have a high level of interpretability for users also). In this paper, we propose a novel fuzzy DBN system called TSK_DBN which combines DBN and TSK fuzzy system. Firstly, the fuzzy clustering algorithm FCM is used to divide the input space, and the membership function of the fuzzy rule is defined. Then, the implicit feature is created by DBN. Finally, the consequent parameters of the fuzzy rule are determined by LLM(Least Learning Machine). The TSK_DBN fuzzy system has an adaptive mechanism, which can automatically adjust the depth until the optimal accuracy is achieved. The prominent character of the TSK_DBN system is that there is adaptive mechanism to regulate the depth of DBN to get a high accuracy. Several benchmark datasets have been used to empirically evaluate the efficiency of the proposed TSK_DBN in handling pattern classification tasks. The results show that the accuracy rates of TSK_DBN are at least comparable (if not superior) to DBN system with distinctive ability in providing explicit knowledge in the form of high interpretable rule base. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1109/ISKE.2017.8258762 | 2017 12th International Conference on Intelligent Systems and Knowledge Engineering (ISKE) |
Keywords | Field | DocType |
TSK fuzzy system,DBN,TSK_DBN,interpretability | Interpretability,Fuzzy clustering,Pattern recognition,Computer science,Deep belief network,Fuzzy logic,Artificial intelligence,Fuzzy control system,Statistical classification,Membership function,Fuzzy rule | Conference |
ISBN | Citations | PageRank |
978-1-5386-1830-1 | 2 | 0.45 |
References | Authors | |
0 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Xiongtao Zhang | 1 | 8 | 2.59 |
| Xingguang Pan | 2 | 3 | 1.16 |
| Shitong Wang | 3 | 1485 | 109.13 |