Algorithm 2.

Learned proximal gradient method.

inputs: measurements y, initial estimate u 0

u ← u 0, s ← 0

for i ← 1, …, it do

$(u,s)\leftarrow {\widehat{\mathrm{prox}}}_i(u,s,\nabla E_y(u))$

end for

return Φ(y) ≔ u