Company Filing History:
Years Active: 2022-2023
Title: Eric Undersander: Innovator in Neural Network Technologies
Introduction
Eric Undersander is a prominent inventor based in San Francisco, CA (US). He has made significant contributions to the field of neural networks, particularly in the areas of model efficiency and audio synthesis. With a total of 2 patents, his work is paving the way for advancements in artificial intelligence and machine learning.
Latest Patents
One of Eric's latest patents is titled "Systems and methods for block-sparse recurrent neural networks." This patent describes innovative systems and methods to prune deep neural network models, effectively reducing their memory and computational requirements. The techniques involve block pruning and group lasso, which allow for the creation of block-sparse recurrent neural networks (RNNs) that maintain accuracy comparable to dense baseline models. By employing these methods, Eric demonstrates that high sparsity can be achieved with minimal loss in performance, enhancing hardware efficiency.
Another notable patent is "Spectrogram to waveform synthesis using convolutional networks." This invention presents an efficient neural network architecture that utilizes transposed convolutions for waveform synthesis from spectrograms. The architecture is designed to achieve high compute intensity and fast inference, significantly improving audio quality. Eric's model can synthesize audio more than 500 times faster than real-time, showcasing its potential for high-quality speech synthesis without relying on iterative algorithms.
Career Highlights
Eric Undersander is currently employed at Baidu USA LLC, where he continues to innovate in the field of artificial intelligence. His work focuses on enhancing the efficiency and effectiveness of neural network models, contributing to the broader landscape of technology and research.
Collaborations
Eric has collaborated with notable colleagues, including Gregory Diamos and Sercan Omer Arik. These partnerships have fostered a collaborative environment that encourages the exchange of ideas and advancements in technology.
Conclusion
Eric Undersander is a key figure in the development of advanced neural network technologies. His patents reflect a commitment to innovation and efficiency in artificial intelligence, making significant strides in both model performance and audio synthesis.