Figure 3.
The parallel coordinate plot visualises model comparison across multiple hyperparameters including batch size, dropout rate, image size, learning rate, and optimisation method. The x-axis lists each parameter, and the y-axis indicates their corresponding values. Lines represent individual models from model_a to model_e, passing through their parameter values. The colour gradient reflects mean average precision, with brighter tones showing higher performance. The process is divided into coarse filtering and fine filtering stages.Illustration of the NAS
Source(s): Created by authors