Adaptive Integrated Layered Attention (AILA) is a neural network architecture that combines dense skip connections with different mechanisms for adaptive feature reuse across network layers.
AILA has been evaluated on price forecasting, image recognition, and sentiment analysis tasks, achieving comparable performance to strong deep learning baselines.
Two versions of AILA have been implemented - AILA-Architecture 1 and AILA-Architecture 2, which differ in their connection mechanisms between layers.
Results show that AILA's adaptive inter-layer connections improve overall performance for various tasks, with reduced training and inference time.