Concept-Oriented Deep Learning: Generative Concept Representations
This work addresses the need for more flexible and integrated concept representations in AI, particularly for domains with complex data types, but it appears incremental as it builds on existing generative models.
The paper tackles the problem of representing concepts in deep learning by proposing generative concept representations, which offer advantages like uncertainty representation and support for learning and reasoning integration, using methods such as variational autoencoders and generative adversarial networks for sequence, structured, or graph data.
Generative concept representations have three major advantages over discriminative ones: they can represent uncertainty, they support integration of learning and reasoning, and they are good for unsupervised and semi-supervised learning. We discuss probabilistic and generative deep learning, which generative concept representations are based on, and the use of variational autoencoders and generative adversarial networks for learning generative concept representations, particularly for concepts whose data are sequences, structured data or graphs.