Transfer learning

What is transfer learning?

Technique that reuses a pre-trained model on a new task. For example, if an image classification model was trained to recognize dogs, the knowledge gained from that task could be applied to try to recognize whales.

Quick start

• Teachable machine

Some Terminology

• Epochs: Iterations. The one entire passing of training data through the machine learning algorithm.
• Activation: a function added into an artificial neural network to help the network learn complex patterns in the data.
• Batch size: Number of training examples used.
• Learning rate: hyperparameter that controls how much the model should change each time the weights are updated.
• Tensor: Mathematical object used to describe physical properties. Generalisation of scalars and vectors.
• Optimizer: a function or an algorithm that adjusts the attributes of the neural network, such as weights and learning rates