Autoregressive sequence models based on deep neural networks, such as RNNs, Wavenet and the Transformer attain state-of-the-art results on many tasks. However, they are difficult to parallelize and are thus slow at processing long sequences. RNNs lack parallelism both during training and decoding, while architectures like WaveNet and Transformer are much more parallelizable during training, yet...

In this paper we investigate the ability of generative adversarial networks (GANs) to synthesize spoofing attacks on modern speaker recognition systems. We first show that samples generated with SampleRNN and WaveNet are unable to fool a CNN-based speaker recognition system. We propose a modification of the Wasserstein GAN objective function to make use of data that is real but not from the cla...

Convolutional autoregressive models have recently demonstrated state-of-the-art performance on a number of generation tasks. While fast, parallel training methods have been crucial for their success, generation is typically implemented in a naı̈ve fashion where redundant computations are unnecessarily repeated. This results in slow generation, making such models infeasible for production environ...

Nowadays, Tibetan speech synthesis based on neural network has become the mainstream method. Among them, griffin-lim vocoder is widely used in because of its relatively simple synthesis.Aiming at problem low fidelity vocoder, this paper uses WaveNet instead for synthesis.This first convolution operation and attention mechanism to extract sequence features.And then linear projection feature ampl...

Background: Underwater images, in general, suffer from low contrast and high color distortions due to the non-uniform attenuation of light as it propagates through water. In addition, degree varies with wavelength, resulting asymmetric traversing colors. Despite prolific works for underwater image restoration (UIR) using deep learning, above asymmetricity has not been addressed respective netwo...