Fig. 2: All-optical intensity linear transformations under spatially incoherent illumination performed by diffractive networks trained using the indirect approach.

All-optical linear transformations of intensity, \(\hat{{\boldsymbol{A}}}\), performed under spatially incoherent illumination, by five of the diffractive network designs shown in Fig. 1c, together with the corresponding error matrices with respect to the target transformation, \({\boldsymbol{\varepsilon }}=\left|{\boldsymbol{A}}-\hat{{\boldsymbol{A}}}\right|\). Here |∙| denotes elementwise operation. The means of the error matrix elements are listed in the table on the right