Skip to main content
. 2026 Jan 9;5:1733003. doi: 10.3389/fradi.2025.1733003

Figure 3.

Diagram illustrating a neural network architecture. On the left is an input chart labeled \"Hydrophobicity\" with values fluctuating over a range. This input is processed through three convolutional layers, each with kernels labeled \\(k_1 \\times k_1\\), \\(k_2 \\times k_2\\), and \\(k_3 \\times k_3\\), maintaining the same padding. Each layer has a set number of channels: \\(n_1\\), \\(n_2\\), \\(n_3\\). The output is then flattened and passed through a fully connected layer with various nodes labeled Alpha, Beta, Delta, Gamma, and Omicron.

A deep learning model processing time-series data. CNN architectures are structured pipelines of layers designed to analyze visual data by using a feature extractor (convolutional, pooling, and activation layers) followed by a classifier (fully connected layers). Key components include convolutional layers for feature extraction, pooling layers to reduce spatial dimensions, and fully connected layers to perform the final classification.