After successes at image classification, segmentation is the next step towards image understanding for neural networks. We propose a convolutional network architecture that includes innovative elements, such as multiple output maps, suitable loss functions, supervised pretraining, multiscale inputs, reused outputs, and pairwise class location filters. Experiments on three data sets show that ou...

This article analyses navigation and language parsing as two instances of the same abstract computation, and suggests that the tool needed may have evolved to serve the former task, and was then reused for the latter. Supporting evidence for the idea, based on the authors’ concept of ‘songline’ navigation, is discussed in the context of current linguistic, psychological and neuroscience researc...

After successes at image classification, segmentation is the next step towards image understanding for neural networks. We propose a convolutional network architecture that includes innovative elements, such as multiple output maps, suitable loss functions, supervised pretraining, multiscale inputs, reused outputs, and pairwise class location filters. Experiments on three data sets show that ou...

Texts propagate among participants in many social networks and provide evidence for network structure. We describe intrinsic and extrinsic evaluations for algorithms that detect clusters of reused passages embedded within longer documents in large collections. We explore applications of these approaches to two case studies: the culture of free reprinting in the nineteenth-century United States ...

In this short technical note we show how PADL – the process algebraic architectural description language of Bernardo, Ciancarini and Donatiello – can be used to specify AFrames. AFrames exist to structure the machine in a Problem Frames development, and are important in that framework as they allow architectural expertise to be captured and reused therein. Because of the close proximity of PADL...