Description
Target setting: Set the target sound absorption coefficient in different frequency ranges, for example: 0.8 in the range of 500Hz to 2000Hz.
GAN training: The initial structural parameters are generated using random noise, and the parameters that match the target are generated by the adversarial training of GAN.
Optimization and evaluation: The generated structural parameters are evaluated, and the optimization algorithm is used to further optimize, and the optimal parameter combination is finally output.
Experimental results and analysis
Prediction model effect: Show the effect of DNN prediction model, including training error, validation error, and test error, and plot the predicted value compared with the true value.
Reverse design effect: Show the structural parameters obtained by GAN reverse design and the corresponding sound absorption performance, and compare with the target sound absorption performance.
Originally posted by @jameswa123 in #53 (comment)