Task scheduling is one key issue in computational overlay grid. For finding solution to this problem, we propose a Chaotic Neural Network Algorithm for Task Scheduling, the goal of this algorithm is to minimize the total execution time of task by appropriately mapping the task to processor in the parallel system. We formulate the task scheduling problem that considers heterogeneousness. To evaluate the effectiveness of this algorithm, we compare this algorithm with the simulated annealing, and the genetic algorithm etc. The average experimental result of the algorithm is obviously superior to other conventional algorithms.