Neural Networks have shown promising performance in various tasks including classification, regression and clustering. Performance of NNs depends on the underlying geometry they designed on. However, most NNs are designed while they are unaware of the data geometry. We aim at tackling this problem and developing new NNs based on recent findings in geometry to learn from complex data.
- Excellent programming skill especially with Python/PyTorch.
- Excellent knowledge and experience with machine learning especially neural networks.