Which neural network type is designed to highlight patterns in input data through convolutions?

• A. Recurrent Neural Networks
• B. Convolutional Neural Networks
• C. Transformers
• D. Generative Adversarial Network

Answer: (B)

Convolutional neural networks are built to detect patterns in input data by applying learnable kernels that slide over the input, creating feature maps that respond to specific patterns such as edges, textures, and shapes. This local pattern detection, combined with weight sharing across the input and hierarchical stacking of layers, lets the network capture spatial structures and invariances, which is especially powerful for images and other grid-like data.

Recurrent networks focus on sequences and temporal dependencies, not on local pattern detection via convolutions. Transformers use self-attention to model relationships across the entire input rather than rely on convolutional pattern highlighting. Generative Adversarial Networks are a framework for training a generator and a discriminator to produce realistic data, not specifically defined by convolution-based pattern highlighting.