Numerical solutions of fuzzy differential equations using artificial neural networks

Many real-world phenomena with uncertainty can be described as fuzzy differential equations. In this paper, we are inspired by a neural network approach toapproximate solutions of linear fuzzy differential equations of first order with initial value under Hukuhara derivative. Firstly, we replace the equations with initial value by systems of linear ordinary differential equations. Secondly, we solve the systems by single-layer feedforward sigmoid-weighted neural networks. We also demonstrate our method with numerical examples which are extremely accurate compared to the exact solutions.