Hi, The material on your web is really good BUT the UI sucks. The combination on colors and font is just not make sense. Please consider re-design the website. Lots of free web template are out there you may want to use one. I could even read . Thanks

hello...my name vita...i'd like to ask about kernel... in some paper i found that the estimaty density of kernel is define by f(x) = 1/n sigma [(x-Xi)/h] but in other paper i found f(x) = 1/(nh) sigma [(x-Xi)/h] what is the differencess...can you help me... and one more question...what is x on that formula...Xi is the data right ? thank you...

Hello Dr. it is very informative site for new researcher like me bcz v are having the basic concepts of the various topics related to research. it would be more interesting and worth full if u will make the video tutorials for all. thanks

I found your notes very useful. For the first time I was able to comprehend how Q-learning works especially the table update. Good work

Hi Dear. i have one question, that is: when we normalize negative numbers, we can use the function that you mentioned at the part 1 of the "normalization methods" on: http://people.revoledu.com/kardi/tutorial/Similarity/Normalization.html now, i want to know what is the mathematical name 0f this equation written, exactly? delta=(1/2)*[1-(d/(d^2+a)] as we know for example: the Laplas equation, the Navie-Stocks equation, the euler equation, etc. thank you for the help and your kindness A. Dolatshah

Hi, my name Steven.. May I ask something? What is the best similarity method to be used for journal plagiarism detection? Cosine or Euclidean or Jaccard or Mahalanobis or Minkowski or anything else that might you suggest..

Hello, We have images of all the number from 0 to 9 in different format,orientation, shape. For example for 0 , we have many images in different, orientation , shape, format but all belonga to same class i.e digit 0 Same for other nmbers. Now using these images we train our model by using feature extraction(regionprops funcion) for every numbers. And then in test image(any image containing charactors and numbers). We will segment each number in this test image so all the numbers are seperated. We only need to recognise numbers, not charactor. So we have class model from trained images. Now for testing image we have to use nearest class mean cluster technique in MATLAB. If in testing image labeled object(particulur number) has feature similar to the images of any class, then that class assigned to that labeled object in tesing image. This is idea. But i'm not able to know which feature to use for every number. IF you can find information about it please let me know. In matlab regionprops function gives you information. And for 0, 1, 2 , 3, 4, 5, 6, 7, 8, 9 numbers we have to define feature vector which classify each numbers from others. Plz give me some informaion. Thanks in Advance.

i read ur tutorial , was nice and keep going, which leads u to be a good teacher

In the matlab code of Q-learning example, it does not reach the goal state in each episode (it stops after updating one action, and continues to next episode). It contradicts the numerical example you gave before. As for the permutation, I think using a uniform random number is sufficient since you only need the first element of the permutation. To generate a uniform random integer from 1 to N in matlab: floor(1+rand(1)*N) Thanks!

Dear Sir/ Good morning, please i want to measure the similarity (dissimilarity) between two binary vectors. is it possible that i use hamming distance and jaccard index (combinaison) and take the best value that can help me to distinguish if the two vectors are different or similar (thresholding). yours faithfully. walid aydi

thank you for the response. i have seen the tutorials it's very usefull. As we know jaccard index is defined as the size of the intersection divided by the size of the union of the sample sets. for me the jaccard distance is more suitable to measure between objects within a cluster and hamming distance is more suitable between objects within different clusters ( i deduced that from the distribution of the intra and inter class). so my question is it possible that i use hamming and jaccard both (combinaison) to compute the distribution of inter and intra class distribution. yours faithfully

Dear Kardi Teknomo, I'm confuse with some journal publication that said they are able to calculate the mahalanobis distance between 2 face images. However, in your tutorial, Mahalanobis is for calculating distance between 2 groups of data. May I know, how to implement mahalanobis distance for 2 data, instead of 2 groups of data? Is this possible? Regards, Jiashen

Hello thankyou so much for these tutorials. I desperately need help on Locality Sensitive Hashing. Any simple way to understand it will be really appreciated. thanks pungu..

Thank you for the excellent excel tutorials! The ideas are so much more accessible with examples.