Applications of Singularity Theory on Deep Neural Network.
(Fapesp process number: 2019/08939-0)
The purpose of this project is to determine singularities on deep neural networks using the math
machinery from Singularity Theory and Applied Mathematics. Such singularities exist due to issues
in the training data, training algorithm, network architecture, algorithm parameters, etc. Singularities
can result in adversarial inputs that can destabilize the output of the network; singularities can also
cause instability in the training process.
Brazil-US SPRINT Project
The SPRINT - São Paulo Researchers in International Collaboration, a strategy designed to encourage and promote the advancement of scientific research through the engagement of researchers affiliated with research institutions in the State of São Paulo with researcher partners abroad.