Do I understand you correctly, that you say cosine distance is a better measurement of similarity than euclidian distance when there are many dimensions?