Transfer Learning Using Deep Neural Networks
Deep learning models now provide state-of-the-art solutions for complex, real world problems. Software for training and testing of deep models is also easily accessible via several libraries. However, training these models from scratch requires huge labeled datasets and heavy computational power, which can pose a severe limitation to their use in practice. Transfer learning offers a way to overcome these limitations. In this technique, knowledge learnt from previously trained models can be applied to solve problems in a related, smaller dataset. This workshop will introduce the basic concepts of transfer learning and provide the know-how to apply it towards developing custom classifiers on real-world datasets.