Fig. 3. Experimental results of CVOCA with input data X_a.

a Four convolution kernels W¹¹, W¹², W²¹, and W¹² were designed based on a real kernel S and its transpose S^T: W¹¹ = S + j·S^T, W¹² = S + j·S, W²¹ = S^T + j·S^T, W²² = S^T + j·S. b The optical spectrum shows the shaped comb for kernel W¹¹, with solid and dashed lines corresponding to positive and negative weights, and dots and stars denoting measured and desired weights. c The designed input data based on a real kernel S and its transpose S^T. d The electrical waveforms generated from the input data. e The output feature of complex-valued convolution between the input data and kernel W¹¹. f The electrical waveforms show the convolution results, with the experimental and simulated results denoted as solid and dashed lines, respectively. The feature maps are experimental results for kernel W¹¹, featuring distinct locations of maximum real and imaginary elements, as highlighted in dashed boxes.