This paper does not exist!
"The Multimodal Recurrent Network: A New Class of Fully Connected Recurrent Networks"
"The multimodal recurrent network is a recently introduced approach to recurrent networks, which can be thought as a hybrid of multilinear and multi-level recurrent networks. The network can be thought as a multimodal recurrent network, and it is able to learn complex nonlinearities such as the inverse normalized logarithmic ratio. As the network is multilinear, it is able to learn nonlinearities that are dependency-free and dependent on only a single input. Moreover, the network is able to learn complex nonlinearities that are to some extent relaxed, and it can be thought as a multi-level recurrent network. The network can be thought as a multilinear recurrent network. The network is multimodal in that the inputs are of any class, and it inherits features such as the inverse normalized logarithmic ratio. The network is multi-level in that the input is of any class, and it can be thought as a multi-level recurrent network. The network is an iterative recurrent network where the inputs are of any class, and it inherits features such as the iterate norm. We evaluate the multicore-based for the multicore-based multimodal recurrent network on a subset of synthetic and real data sets. We obtain the best performance on synthetic and real data sets."