Primitive Recursion Without Composition: A Dynamic Equivalence Characterization from Neural Networks to Polynomial ODEs
A new arXiv paper proves that recurrent ReLU neural networks, polynomial ordinary differential equations, and discrete p…
📂 Research
Latest articles in Research
Learn&Drop: A Novel Layer-Dropping Method for Faster CNN Training
A latest arXiv paper proposes the Learn&Drop method, which dramatically improves deep convolutional neural network train…
SeaEvo: Advancing Automatic Algorithm Discovery Through Strategy Space Evolution
A research team has proposed the SeaEvo framework, which introduces an explicit strategy space evolution mechanism to ad…
New PINN Framework Simulates Arctic Pollution Dispersion
Researchers propose a robust physics-informed neural network framework based on the collocation method to simulate time-…
Why Architecture Choice Matters in Symbolic Regression
A recent arXiv paper explores how operation tree structures in symbolic regression affect formula discovery capabilities…
AeSlides: Using Verifiable Rewards to Help AI Create More Visually Appealing Presentations
A latest arXiv paper introduces the AeSlides framework, which uses a verifiable reward mechanism to incentivize large la…
Multiplication-Free Spiking Neural Network Learning Algorithm Enables Efficient FPGA On-Chip Training
Researchers propose a multiplication-free spike-time learning algorithm designed specifically for efficient FPGA impleme…
GEGLU-Transformer Enables IMU-to-EMG Signal Estimation
Researchers propose a GEGLU-Transformer-based adaptive learning framework that reconstructs continuous muscle activation…
Semantic Relationship-Enhanced Graph Matching Improves SLAM Localization Robustness
A new study proposes a semantic-enhanced graph matching method that explicitly models semantic relationships between obj…
ANCHOR Framework: Making Home Service Robots More Robust
A research team has proposed the ANCHOR framework, which addresses the inconsistency between symbolic planning and the p…
ProDrive: A New Paradigm for Proactive Planning in Autonomous Driving
A research team has proposed the ProDrive framework, which achieves proactive planning through a world model that co-evo…
New Learning-Based Framework for Modeling and Control of Tendon-Driven Continuum Robots
A latest arXiv paper proposes a differentiable learning framework that combines high-fidelity dynamics modeling with rob…