About this Research Topic
In recent years, neural network research and applications have received renewed attention fostered by the advances in representation learning and the idea of ‘Deep Learning’. Enabled by technological progress in information technologies, specialized hardware and the availability of Big Data, new algorithms, architectures, and methods have been introduced that allow the practical estimation of neural networks with many hidden layers and diverse configurations. As a consequence, this revolutionized both academia and industry. Key elements for their success were Restricted Boltzmann machines and backpropagation in order to learn different types of neural networks, e.g., deep belief networks (DBN), convolutional neural networks (CNN) and long short-term memory (LSTM) networks.
Despite all these successes, we are still at the beginning to fully understand deep neural networks theoretically and how we could apply them efficiently to domain-specific data. For instance, so far the most successful application breakthroughs have been achieved for domains as image and speech analyses. However, there are many more data types that are waiting to be fully explored and present different characteristics. Especially in computational social science, there is a huge potential for the application of such methods. Examples for fields from which such data could come from that are beyond image and audio data are business, economy, finance, human behavior, management, public health, social media, and psychology.
A particular data type for which deep learning holds great promises is text data, which can be the basis for many prediction or pattern detection tasks relevant in social sciences and human problems. For such data, novel methodological developments, as well as new application areas, are of interest.
The purpose of our research topic ‘Deep Learning in Computational Social Science’ is to bring together researchers from university and industry interested in this interdisciplinary topic to share information about recent progress and persisting challenges. We are particularly interested in accommodating contributions that explore any of the following research topics (but not limited to):
I. Mathematical models:
• stability of prediction models
• classification of networks
• benchmarking of deep learning methods using social data
• structure learning with neural networks in human and social domains
• representation learning and interpretable structures
• novel unsupervised and supervised learning methods
• statistical requirements for supervised models
II. Applications:
• applications in business analytics and industrial processes
• applications in finance
• applications in marketing
• applications in public health
• applications in natural language processing
• applications in psychology
• application in social media
Keywords: deep learning, big data, neural networks, social science, human behaviour
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.