Modeling the Dynamics of Multimodal Discrete Event-based Multimedia Information Using Recurrent Neural Networks – We propose a recurrent neural network (RNN) architecture for the video-based video retrieval task. We present two versions of the proposed recurrent network with two different feature matrices: one that represents the temporal context as a smooth image, and the other that represents a sequence of the same temporal context. The proposed architecture achieves state-of-the-art accuracy on all video retrieval tasks — including those involving videos of human action and character recognition. The model consists of two stages: the training stage, consisting of 1) a network trained to encode the temporal context as a smooth image and 2) a recurrent network to model the motion as a sequence of high-level visual concepts. The proposed architecture is able to handle the non-linearity and non-linearity within our temporal context models. Moreover, it is also able to integrate temporal and non-linearity and represent both simultaneously. Experimental results on video retrieval data show that our model achieves comparable or better accuracy than state-of-the-art recurrent neural networks.
We present a new method for improving human performance due to the use of high-level features extracted from linguistic resources. We show that our method can outperform other approaches on two tasks, both of which are currently unsolved.
Estimating Linear Treatment-Control Variates from the Basis Function
Distributed Optimistic Sampling for Population Genetics
Modeling the Dynamics of Multimodal Discrete Event-based Multimedia Information Using Recurrent Neural Networks
A Bayesian Model for Data Completion and Relevance with Structured Variable Elimination
Towards a more balanced model of language acquisitionWe present a new method for improving human performance due to the use of high-level features extracted from linguistic resources. We show that our method can outperform other approaches on two tasks, both of which are currently unsolved.
Leave a Reply