We attempt to recover an unknown function from noisy, sampled data. Using o
rthonormal bases of compactly supported wavelets, we develop a nonlinear me
thod which works in the wavelet domain by simple nonlinear shrinkage of the
empirical wavelet coefficient. The shrinkage can be tuned to be nearly min
imax over any member of a wide range of Triebel- and Besov-type smoothness
constraints and asymptotically minimax over Besov bodies with p less than o
r equal to q. Linear estimates cannot achieve even the minimax rates over T
riebel and Besov classes with p < 2, so the method can significantly outper
form every linear method (e.g., kernel, smoothing spline, sieve) in a minim
ax sense. Variants of our method based on simple threshold nonlinear estima
tors are nearly minimax. Our method possesses the interpretation of spatial
adaptivity; it reconstructs using a kernel which may vary in shape and ban
dwidth from point to point, depending on the data. Least favorable distribu
tions for certain of the Triebel and Besov scales generate objects with spa
rse wavelet transforms. Many real objects have similarly sparse transforms,
which suggests that these minimax results are relevant for practical probl
ems. Sequels to this paper, which was first drafted in November 1990, discu
ss practical implementation, spatial adaptation properties, universal near
minimaxity and applications to inverse problems.