Doctoral Candidate Tasha Nagamine Wins Grant to Present Two Papers at Interspeech 2016

Tasha Nagemine has won a grant to attend the conference Interspeech 2016 in San Francisco California where she will be presenting two papers. The grant will fund her registration for the conference, her membership to ISCA, and travel.

The first paper that Tasha will be presenting at “Interspeech: Understanding Speech Processing in Humans and Machines.” is titled “Adaptation of neural networks constrained by prior statistics of node co-activations," and proposes an unsupervised algorithm for introducing adaptive feedback into neural networks models.  This idea is inspired by biological neural networks, which use feedback to dynamically adapt their computation when faced with unexpected inputs.  The second, “On the Role of Nonlinear Transformations in Deep Neural Network Acoustic Models,” attempts to shed light into the “black box” of neural network models in the realm of speech recognition by exploring how the network nonlinearities contribute to robust representations of speech. 

Tasha is a PhD candidate in the electrical engineering department working with Nima Mesgarani.  She graduated from Brown University with a B.S. in physics and began her degree in EE in 2014.  Her current research interests are focused on deep neural networks models for speech and language processing.  She is currently supported by the From Data to Solutions NSF IGERT fellowship.