The topic of this KNIME meetup is codeless deep learning. The first talk by Kathrin Melcher gives you an introduction to recurrent neural networks and LSTM units followed by some example applications for language modeling. In the second talk, Corey Weisinger will present the concept of transfer learning. He’ll be discussing a use case where the VGG16 image classification model is applied to cancer cell diagnosis.
Introduction to RNN and their Applications - Kathrin Melcher
RNN units in deep learning architectures are the state of the art for sequence analysis. After introducing recurrent neural networks and LSTM units, we will take a look at different sequence based use cases, using many to one, many to many, and one to many neural architectures. Let’s find out together how these architectures differ from each other and how they can be used for language modeling.
Transfer Learning for Image Classification - Corey Weisinger
Training deep learning image classifiers requires tons and tons of data, but transfer learning can allow us to exploit models trained on related datasets to reduce this requirement. The VGG16 image classification model is trained on fairly mundane images, but the hope is that the latent features learned inside the network will be helpful when classifying our cancer cells.