Automated Data Science and Deep Neural Network Synthesis - overview
Deep learning techniques have been the key to major improvements in machine learning in various domains such as image and speech recognition and machine translation. Traditionally, people built handcrafted neural network architectures for a particular task or they manually adapt popular networks, such as ResNet, DenseNet, Inception network, to the task at hand. As finding an architecture manually is extremely arduous and requires an exploration of several network architectures, there has been an increased effort in automating it.
The scope of this internship is push the boundaries in the domain of automated machine learning in particular with respect to deep learning. This could include data preprocessing, data augmentation, neural architecture search, automatic ensembling or similar tasks of interest.
Upon successful completion, this internship will have helped to develop principled methodologies that advances the field of AutoML in general. It will have demonstrated the effectiveness of that methodology by systematic experimental evaluation and it will have led to a prototype implementation demonstrating how AutoML can be assessed in a real-world setting.
- Working knowledge in deep neural networks and applications such as image or text classification.
- Strong Python programming skills
- Experience of working with deep learning libraries, e.g. TensorFlow, Pytorch, Keras
- Solid understanding of mathematical concepts such as (Bayesian) Statistics and Linear Algebra.