Instance-based Learning

• an object category is represented by a set of known instances a nearest neighbor classifier is used
• IBL considers category learning as a process of learning about the instances of the category:
• The training phase is very fast
• IBL can recognize objects using a very small number of experiences IBL is a baseline approach to evaluate object representations Simple and easy to implement
• Memory usage in instance-based systems is continuously growing. Computational complexity grows with the number of training instances
• The computational complexity of classifying a single new instance is O(n), where n is number of instances stored in perceptual memory.
• Salience and forgetting mechanisms can be used to bound the memory usage which are also useful for reducing the risk of overfitting to noise in the training set.
• Overfitting
• Sensitive to noise
•