4 ms·
Not really. Given an estimate of the mean; for each data point, Mean shift defines a window around it and computes a new estimated mean weighting each point by
by bazzargh 10y ago
Not really.
Given an estimate of the mean; for each data point, Mean shift defines a window around it and computes a new estimated mean weighting each point by the probability density at the previous estimated mean calculated using the window
The (weighted) 'mean of the data points within the window' makes sense if you use the other perspective of looking at the window around the current estimated mean - you'll get the same answer, and to me this explanation is easier to grasp - the PDF only depends on the distance between the point and the estimated mean so you can think of either as the 'center'. But saying you calculate the mean of the data points within the window for each data point mixes up two perspectives and makes no sense.
- ebalit 10y agodaveguy's explanation is correct. It's the simplest form of the Mean Shift algorithm.
- bazzargh 10y agoNope. It would be correct if it said 'For each estimate of the mean, Mean shift defines a window around it and computes the mean of the data points within the window, then repeats the process with this new estimate.' Suppose we had 500 data points. Daveguy's process calculates a separate mean for a window around each data point. Now we have 500 means...and?