search

sunnyMiglani / SPS_Fourier

Fourier Analysis done for handwriting recognition for SPS Module at UoB 2017 with Sydney
0 stars 0 forks source link

Distance measures #12

Open SydneyJake opened 7 years ago

SydneyJake commented 7 years ago

Let's analyse the NN classifier with different distance measures to see if we get a better classification, then try and explain why.

sunnyMiglani commented 7 years ago

What values have you tried till now?

SydneyJake commented 7 years ago

I have tried:

Euclidean (Current) - Will "fairly" take into account both the X and Y

Manhattan - Not sure how to analyse this one, its sort of a mix of the two.

Chebyshev - This will pick NN based on largest X or Y difference from point we want to classify, might be better at classifying outliers.

I'll get diagrams and classification labels for all so we can talk about them in the report. Is there any one that stands out as more suited for our purposes @sunnyMiglani ?