How do CNNs balance depth vs width in architecture design?

CNN design balances depth (feature hierarchy) and width (feature diversity) to optimize performance and computational cost.

Depth increases representational hierarchy, enabling learning of abstract features. Width increases the number of parallel feature detectors, improving diversity. Very deep networks may suffer optimization issues, while very wide networks may become computationally expensive. Modern architectures like ResNet and EfficientNet carefully balance both using scaling laws.