Graph filter with the heat kernel: $$f(x) = exp(-\beta |x / \lambda_m - a|^b)$$
heatFilter(x, l.max, order = 1, offset = 0, beta = 30)
numeric Values to be filtered. Normally, these are graph laplacian engenvalues.
numeric Maximum eigenvalue on the graph (\(\lambda_m\) in the equation)
numeric Parameter \(b\) in the equation. Larger values correspond to the sharper kernel form (default=1). The values should be positive.
numeric Mean kernel value (\(a\) in the equation), must be in [0:1] (default=0)
numeric Parameter \(\beta\) in the equation. Larger values provide stronger smoothing. \(\beta=0\) corresponds to no smoothing (default=30).
smoothed values for `x`
Other graph smoothing:
computeChebyshevCoeffs()
,
smoothChebyshev()
,
smoothSignalOnGraph()