Table 6 Hyperparameter settings for the proposed CNN-based cooling system prediction.
From: Deep regression analysis for enhanced thermal control in photovoltaic energy systems
Parameter | Description | Value |
|---|---|---|
Batch size | Size of batches used during training | 32 |
Epochs | Total number of iterations over the entire dataset | 50 |
Steps per epoch | Number of batches processed during one epoch | Len(X_train)//batch_size |
Validation steps | Number of batches processed during validation | len(X_test)//batch_size |
Learning rate | The initial learning rate for the optimizer | 0.0001 (Adam) |
Decay | Decrease factor for learning rate | 1e-4/256 (Adam) |
Dropout rate | The proportion of units randomly dropped during training | 0.5 (Convolutional layers) |
Num filters | Number of filters in convolutional layers | Varies (Convolutional layers) |
Kernel size | Spatial extent of convolutional kernels | 3 (Convolutional layers); 2 (Upconvolutional blocks) |
Pool size | Downsampling factor in max pooling layer | 2 (Convolutional Blocks) |
Activation function | Mathematical function defining node behavior | ReLu; Softmax |
Loss function | Objective function minimized during optimization | Categorical crossentropy; Mean squared error |
