Jump to content
Image picture of a abstract network

Applied Artificial Intelligence

A central part of Data Science is making predictions from sampled data. Thus, Machine learning - which emerged from the early stages of artificial intelligence and evolved from pattern recognition and computational learning theory – is amongst the core competences of the research group.

Competence and technology

Machine Learning makes use of statistical learning which derives a predictive function based on data. This data can be manifold in terms of quantity, quality and type – including media such as audio, video and text, sensor data, system logs, etc. The core competence of the team lies in the domains of audio analysis, natural language processing, graph analysis as well as multi-modal combinations among them. Though, machine learning is used by many others, the preeminent feature of the Digital Insights Lab is, that it covers the entire data science life-cycle from data ingestion – ranging from small to very large data - preparation, modelling and prediction using machine learning to finally reasoning and conclusions.

Deep Learning

Deep Learning is a machine learning approach using Neural Networks. Several layers are used to make these networks “deep” and different types of layers are used to model specific tasks such as recognizing objects in images, named entities in text or acoustic events in audio recordings. Deep Learning has gained a lot of popularity because it yields remarkable results in various application domains. The advantage of deep learning is, that it entirely learns predictive functions from raw data. This distinguishes it from former approaches which required to extract hand-crafted features from the data. These features required precise knowledge of the problem domain as well as the technical skills to extract/calculate them. Deep learning therefore relieves the machine learning specialist of the need of such specialized skills and facilitates more generalized approaches. Further it facilitates the integration of different data modalities such as combining acoustic, visual and text input in one model. The Digital Insights Lab accomplished to successfully apply deep learning to project relevant tasks and to successfully compete against competing research groups in international evaluation campaigns.