Texture is a visual attribute largely used in many problems of image\nanalysis. Currently, many methods that use learning techniques have been\nproposed for texture discrimination, achieving improved performance over\nprevious handcrafted methods. In this paper, we present a new approach that\ncombines a learning technique and the Complex Network (CN) theory for texture\nanalysis. This method takes advantage of the representation capacity of CN to\nmodel a texture image as a directed network and uses the topological\ninformation of vertices to train a randomized neural network. This neural\nnetwork has a single hidden layer and uses a fast learning algorithm, which is\nable to learn local CN patterns for texture characterization. Thus, we use the\nweighs of the trained neural network to compose a feature vector. These feature\nvectors are evaluated in a classification experiment in four widely used image\ndatabases. Experimental results show a high classification performance of the\nproposed method when compared to other methods, indicating that our approach\ncan be used in many image analysis problems.\n