Hyper-Parameter Optimization of Generative adversarial network (GAN) -- 2
Budget: $30 – $250 USD
Highly sensitive to the hyper parameter selections
The Google Brain paper mentions that GAN is sensitive to hyper parameter optimization due to the benefit of cost function can be overridden from the hyper parameter optimization’ benefit if the can override the cost function cost function is not properly optimized so the performance of the cost functions can fluctuate within the different settings of hyper parameter and this fluctuating can be desperate when the developers don’t know whether the GAN’ model is not working or they need to engage in lengthy [8].
The figure below proves that the training of GAN extremely is sensitive within the different settings of hyper parameter and this indicates that there is no GAN’ model is significantly more stable than others. The Black stars in the figure mention to the suggested hyper parameter settings’ performance with a wide range hyper parameter search (100 hyper parameter samples per model) [5].
The Google Brain paper mentions that GAN is sensitive to hyper parameter optimization due to the benefit of cost function can be overridden from the hyper parameter optimization’ benefit if the can override the cost function cost function is not properly optimized so the performance of the cost functions can fluctuate within the different settings of hyper parameter and this fluctuating can be desperate when the developers don’t know whether the GAN’ model is not working or they need to engage in lengthy [8].
The figure below proves that the training of GAN extremely is sensitive within the different settings of hyper parameter and this indicates that there is no GAN’ model is significantly more stable than others. The Black stars in the figure mention to the suggested hyper parameter settings’ performance with a wide range hyper parameter search (100 hyper parameter samples per model) [5].