NFI: a neuro-fuzzy inference method for transductive reasoning

Qun Song, N.K. Kasabov
2005 IEEE transactions on fuzzy systems  
This paper introduces a novel neural fuzzy inference method-NFI for transductive reasoning systems. NFI develops further some ideas from DENFIS-dynamic neuro-fuzzy inference systems for both online and offline time series prediction tasks. While inductive reasoning is concerned with the development of a model (a function) to approximate data in the whole problem space (induction), and consecutively-using this model to predict output values for a new input vector (deduction), in transductive
more » ... oning systems a local model is developed for every new input vector, based on some closest to this vector data from an existing database (also generated from an existing model). NFI is compared with both inductive connectionist systems (e.g., MLP, DENFIS) and transductive reasoning systems (e.g., K-NN) on three case study prediction/identification problems. The first one is a prediction task on Mackey Glass time series; the second one is a classification on Iris data; and the last one is a real medical decision support problem of estimating the level of renal function of a patient, based on measured clinical parameters for the purpose of their personalised treatment. The case studies have demonstrated better accuracy obtained with the use of the NFI transductive reasoning in comparison with the inductive reasoning systems. Index Terms-Adaptive systems, neural-fuzzy inference (NFI), renal function evaluation, time series prediction , transductive reasoning. I. INDUCTIVE VERSUS TRANSDUCTIVE LEARNING AND REASONING SYSTEMS M OST learning models and systems in artificial intelligence developed and implemented so far [31], especially in the area of soft computing [10], [11], [13], [19], [28], [32], [37], [40], and particularly-in neuro-fuzzy reasoning systems [5], [6], [8], [20], [21], [30] are based on inductive inference methods, where a model (a function) is derived from data representing the problem space and this model is further applied on new data. The model is usually created without taking into account any information about a particular new data vector (test data). An error is measured to estimate how well the new data fits into the model. The models are in most cases global models, covering the whole problem space. Such models are for example: regression functions; the multilayer perceptron neural network (MLP) used in this paper to compare results with, and also-the ANFIS neuro-fuzzy inference system [20] . These models are difficult
doi:10.1109/tfuzz.2005.859311 fatcat:dunje3ne3rgu5ldcytgog7medi