KNN in contrary to other algorithms doesn’t build any model, do you think this is a privilege because it doesn’t depend on any training data so when we enrich the data its performance increases?
Hi Ikram,
Yes, this can be correct; as it doesn't form any assumptions about characteristics of the concepts. In addition, KNN learning cost is very low especially when the data is small. Finally, complex aspects can be learned using simple procedures.
HTH.
Samer
k-NN is also very easy to adapt for data stream and active learning scenarios
The way you define distance between your data points is essentially a form of assumption. The ordinary knn does not rate the relative importance of individual features for discrimination and consider all features as equally important, this can be considered itself as a significant assumption about features.
Knn is actually very sensitive to noise, outliers, size of training sample and value of K.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence