In a traditional machine learning paradigm (see Figure 2.1), every use case or task is modeled independently, based on the data at hand. In transfer learning, we use the knowledge gained from a particular task (in the form of architecture and model parameters) to solve a different (but related) task, as illustrated in the following diagram:
Andrew Ng, in his 2016 NIPS tutorial, stated that transfer learning would be the next big driver of machine learning's commercial success (after supervised learning); this statement grows truer with each passing day. Transfer learning is now used extensively in problems that need to be solved with artificial neural networks. The big question, therefore, is why this is the case.
Training an artificial neural network from scratch is a difficult...
