AIExplainer
Machine Learning Intermediate 2 min read

What is transfer learning?

A machine learning technique where a model trained on one task is used as a starting point for another related task

Transfer learning is a way to use knowledge gained from one problem to help solve another problem. It's like using a map of one city to help navigate a similar city, instead of starting from scratch

A person who knows how to ride a bike can more easily learn to ride a scooter, because they already understand how to balance and steer. Similarly, a model trained on one task can be fine-tuned for another task, using the knowledge it already has

Google's image recognition model was trained on a large dataset of images, and then fine-tuned for specific tasks like recognizing dogs or cats. This allowed the model to learn from the general knowledge it gained from the large dataset, and apply it to the specific task

Transfer learning is used in many areas, such as image recognition, natural language processing, and speech recognition. It's especially useful when there's not enough data available for the new task, or when the new task is similar to the original task

Some people think that transfer learning means simply copying a model from one task to another, but it actually involves fine-tuning the model to fit the new task. Others think that transfer learning only works for very similar tasks, but it can be used for a wide range of tasks

Transfer learning has been around since the 1990s, but it gained popularity in the 2010s with the development of deep learning models. It's now a key technique in many areas of AI research

domain adaptation knowledge transfer multi-task learning

Three products for different needs — explore what’s relevant to you.