From the course: AI Solution Design Patterns: Data, Model Training, and Application Architectures

Unlock this course with a free trial

Join today to access over 24,600 courses taught by industry experts.

Transfer learning

Transfer learning

From the course: AI Solution Design Patterns: Data, Model Training, and Application Architectures

Start my 1-month free trial Buy for my team

Transfer learning

“

- The transfer learning pattern addresses the problem of wasting the time, energy, and cost of training a model on a task that is similar to tasks for which other models already exist. So let's imagine we need to train a model from scratch using a massive amount of training data. For example, a model that needs to be trained on knowledge in a certain field of medicine. This type of large scale training effort will take a long time and will require a significant financial investment. Now let's imagine that there's a model that's already been trained on a lot of this medical knowledge. If we could reuse that model as the starting point for our new model, we dramatically reduce the scope of the training effort. That is what transfer learning is. Instead of training the model from scratch, we start with a pre-trained model and then further train that model to customize it the way we need to using additional training data. Transfer learning is a powerful way to accelerate model training by…

Contents