I want to know how can I choose distance or similarity measure for my data.
for example when euclidean distance is good or when cosine measure is better than that?
I think it is data dependent. Altough the cosine measure is commonly used as a semantic measure between two feature vectors, they do not always perform perfectly.
if you are comparing text is different than comparing image data. May be ecludian can work better in cases where data is numeric and has a semantic meaning by distance.
have a look at the attached paper.
Issa
Chapter Joint Distance and Information Content Word Similarity Measure
Similarity measures can be compared in terms of simplicity, speed, dimnesionality, immunity to outliers and noise, etc.
I found a new metric which is invariant to data scale, noise and outliers, simple and fast.
It is perfect for CBIR systems, a student of mine is making excellent results using this distance, you may find the paper at:
https://www.researchgate.net/publication/264995324_Dimensionality_Invariant_Similarity_Measure
Article Dimensionality Invariant Similarity Measure
Assalamualaikum / Hai Faezah
My situation is same as you. I try to study all the similarity method, and I realize that the MOST IMPORTANT is looked at your data/ application strength. Where to apply / use your data. Then, It is not wrong to choose the suitable one depended on our literature.
Ok, just a sharing experience.
Thank you, goog luck.
- Badges
- Science topic