Discovering the Arrow of Time in Machine Learning

Machine learning (ML) is increasingly useful as data grow in volume and accessibility. ML can perform tasks (e.g., categorisation, decision making, anomaly detection, etc.) through experience without explicit instruction, even when the are too vast, complex, highly variable, full of errors to be analysed other ways. Thus, great for natural language, images, or complex messy available large growing volumes. Selecting models depends on many factors they vary supervision needed, tolerable error levels, ability account order temporal context, among things. Importantly, methods that use explicitly ordered time-dependent struggle with asymmetry. Most (implicitly) time-dependent, potentially allowing a hidden `arrow time’ affect performance non-temporal tasks. This research explores interaction implicit using two automatically classify (a task) tweets (temporal data) under conditions balance complexity data. Results show was affected, suggesting researchers should carefully consider time matching appropriate tasks, only implicitly included.