KEYWORD |
Implementation of Recurrent Neural Networks on embedded devices
keywords NEURAL NETWORKS, EMBEDDED SYSTEMS, C++ PROGRAMMING
Reference persons MASSIMO PONCINO
Research Groups GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA
Thesis type MASTER THESIS
Description While for traditional convolutional neural networks (CNN) many hardware implementations do exist (for platforms of different computational power), the landscape for recurrent networks (RNN) is much poorer and solutions are only avaiable for devices with medium- to -high computational power.
The objective of this thesis is that of implementing a working prototype of a RNN on an embedded devices (to be defined as part of the project) to validate existing RNN solutions/algorithms on text/voice processing/recognition.
An essential part of the project is to define suitable approximations of the RNN in order to fit the
network to the reduced SW/HW capabilities of the device.
Required skills C/C++ programming, embedded systems programming
Deadline 01/05/2019
PROPONI LA TUA CANDIDATURA