Deep Learning: revealing the "known unknowns" in the archaeological record
Deep learning is a subfield of machine learning and artificial intelligence (AI) that uses algorithms to drive the exploration of "known unknowns" hidden in large datasets. Deep learning algorithms are inspired by connectivity patterns in the brain’s visual cortex that mimic the ability to learn hierarchies of concepts and building up multiple layers of abstraction. These form the basis for convolutional neural networks (CNNs).
We rely on multi-disciplinary investigations involving archaeologists, anthropologists, chemists and computer scientists. We focus on acquisition of large archaeology dataset and development and validation of suitable CNNs-based algorithms to infer “hidden patterns”. It yields new and testable working hypotheses for artifact identification, classification, and geographic correlations.