 
Summary: Monte Carlo Algorithm for Least Dependent
NonNegative Mixture Decomposition
Sergey A. Astakhov,*, Harald Sto1gbauer,, Alexander Kraskov,, and Peter Grassberger
John von Neumann Institute for Computing, Forschungszentrum Ju¨lich, D52425, Ju¨lich, Germany, and Division of Biology,
California Institute of Technology, Pasadena, California 91125
We propose a simulated annealing algorithm (stochastic
nonnegative independent component analysis, SNICA)
for blind decomposition of linear mixtures of nonnegative
sources with nonnegative coefficients. The demixing is
based on a Metropolistype Monte Carlo search for least
dependent components, with the mutual information
between recovered components as a cost function and
their nonnegativity as a hard constraint. Elementary
moves are shears in twodimensional subspaces and
rotations in threedimensional subspaces. The algorithm
is geared at decomposing signals whose probability densi
ties peak at zero, the case typical in analytical spectro
scopy and multivariate curve resolution. The decomposi
tion performance on large samples of synthetic mixtures
and experimental data is much better than that of tradi
