Temporal and Object Quantification Networks
This addresses the challenge of learning relational-temporal patterns in neuro-symbolic AI, with incremental improvements in generalization to unseen object counts and temporal variations.
The paper tackles the problem of recognizing complex relational-temporal events by introducing Temporal and Object Quantification Networks (TOQ-Nets), which generalize to varying numbers of objects and temporal sequence lengths, as demonstrated through evaluation on event-type recognition tasks.
We present Temporal and Object Quantification Networks (TOQ-Nets), a new class of neuro-symbolic networks with a structural bias that enables them to learn to recognize complex relational-temporal events. This is done by including reasoning layers that implement finite-domain quantification over objects and time. The structure allows them to generalize directly to input instances with varying numbers of objects in temporal sequences of varying lengths. We evaluate TOQ-Nets on input domains that require recognizing event-types in terms of complex temporal relational patterns. We demonstrate that TOQ-Nets can generalize from small amounts of data to scenarios containing more objects than were present during training and to temporal warpings of input sequences.