Abstract
Relying simply on bitwise operators, the recently introduced Tsetlin machine (TM) has provided competitive pattern classification accuracy in several benchmarks, including text understanding. In this paper, we introduce the regression Tsetlin machine (RTM), a new class of TMs designed for continuous input and output, targeting nonlinear regression problems. In all brevity, we convert continuous input into a binary representation based on thresholding, and transform the propositional formula formed by the TM into an aggregated continuous output. Our empirical comparison of the RTM with state-of-the-art regression techniques reveals either superior or on par performance on five datasets. This article is part of the theme issue 'Harmonizing energy-autonomous computing and intelligence'.
Highlights
Over the last few years, there has been tremendous progress in research on predictive models, resulting in increasingly higher predictive accuracy
Tsetlin machine (RTM), which consists of three parts: (i) For continuous input, we propose a data preprocessing procedure that transforms the input losslessly into a binary representation that maintains semantic relationship between numbers
We study the behaviour of the regression Tsetlin machine (RTM) on five different real-world datasets: Dengue Incidences:3 This dataset consists of monthly dengue incidences in the Philippines per 100 000 population
Summary
Over the last few years, there has been tremendous progress in research on predictive models, resulting in increasingly higher predictive accuracy. The Tsetlin machine (TM) is a recent pattern recognition approach that attempts to bridge the gap between the interpretability of rule-based techniques and the high predictive accuracy of deep learning [10]. The TM is relatively simple computationally, being based on propositional logic to form the classifier, and straightforward bitwise operations for recognition and learning This structure makes the TM interpretable, yet it achieves competitive accuracy for many pattern recognition problems. Continuous input and output can be encoded in bit form, the natural ordering of numbers is lost We address this limitation in the present paper by introducing the regression. (iii) we propose a new feedback scheme for guiding the TA of the RTM, to support regression This scheme minimizes the discrepancy between the predicted and the target outputs, with the aid of a modified stochastic clause activation function.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call