AdaKAN: New KAN-based network advances medical image segmentation

Researchers have introduced AdaKAN, a novel neural network designed for medical image segmentation. This model integrates convolutional operations with an efficient Kolmogorov-Arnold Network (KAN) block, featuring an adaptive KAN module with a dual-branch design. One branch uses KAN layers with Bernstein polynomial activations for smooth function approximation, while the other refines features through projection and adaptive scaling. AdaKAN employs a U-shaped architecture to capture both long-range and local dependencies, outperforming conventional convolutional and Transformer models in segmentation accuracy across various medical imaging datasets. AI

IMPACT AdaKAN's novel architecture may offer improved accuracy and efficiency for medical image segmentation tasks.

RANK_REASON The cluster contains a research paper detailing a new model for a specific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• AdaptKAN
• arXiv
• Bernstein polynomial
• EffiKAN
• Kolmogorov--Arnold Networks
• medical image segmentation
• Transformer

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →