The Gated Multimodal Unit (GMU) is a model developed to be used as an internal unit within a neural network architecture. GMU aims to learn an intermediate representation through multimodal information fusion; the GMU learns the relative importance of each modality and combines them to produce a common representation. This unit takes advantage of multimodal information that describes the same phenomena through different sensors, taking complementary and common information to produce better representations.
Victor Hugo Contreras Ordoñez
Jhon Edilson Arevalo Ovalle (Ex-Miembro)
Rosso-Mateus, A., González, F. A., & Montes-y-Gómez, M. (2017, September). A Shallow Convolutional Neural Network Architecture for Open Domain Question Answering. In Colombian Conference on Computing (pp. 485-494). Springer, Cham. doi: 10.1007/978-3-319-66562-7_35. (PDF).
Yao, L., Zhang, Y., Feng, Y., Zhao, D., & Yan, R. (2017). Towards implicit content-introducing for generative short-text conversation systems. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing (pp. 2190-2199).
Yan, R., & Zhao, D. (2018, July). Smarter Response with Proactive Suggestion: A New Generative Neural Conversation Paradigm. In IJCAI (pp. 4525-4531).
Gammulle, H., Denman, S., Sridharan, S., & Fookes, C. (2018). Multi-Level Sequence GAN for Group Activity Recognition. arXiv preprint arXiv:1812.07124.