IMAINET: An Immune Algorithm-Inspired Neural Network Framework

<p dir="ltr"><b>IMAINET </b>is a self-organizing neural network framework inspired by immune algorithms such as clonal selection and affinity maturation. The architecture features a two-phase learning process: the first phase self-organizes hidden units based on immune principles, while the second phase learns the output mapping using various optimization strategies.</p><p dir="ltr"><b>IMAINET </b>is built with <i>PyTorch </i>for flexible gradient-based learning and supports metaheuristic algorithms via <i>Mealpy</i>, enabling robust optimization of network weights. Additionally, it offers an option for closed-form training using <i>least squares estimation</i> (e.g., ridge regression).</p><p dir="ltr">Wrapped in Scikit-Learn's <b>BaseEstimator</b>, <b>IMAINET </b>is easy to integrate into existing ML workflows, supporting pipelines, cross-validation, and hyperparameter tuning.</p>