Distributed representation is a method in machine learning where information is represented as a pattern across multiple nodes or units, allowing for more efficient and nuanced encoding of data. This approach enables models to generalize better by capturing complex relationships and dependencies within the data.