Manchester Literary and Philosophical Society
PulseAugur coverage of Manchester Literary and Philosophical Society — every cluster mentioning Manchester Literary and Philosophical Society across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
MetaNCA learns to self-organize neural network weights
Researchers have introduced Meta Neural Cellular Automata (MetaNCA), a novel framework designed to learn local rules that can self-organize the weights of artificial neural networks. This approach utilizes a learned rul…
-
Neural network symmetry readout depends on positional encoding
A new paper explores how the symmetries within neural network weights are influenced by the positional encoding (PE) and the specific readout methods used for analysis. Researchers found that the PE can obscure or revea…
-
Spiking Neural Network Achieves In-Context Learning with Single Layer
Researchers have developed DendriCL, a novel single-layer spiking neural network architecture that demonstrates in-context learning (ICL) capabilities. Unlike existing AI models that rely on deep architectures and impli…
-
New S-GAI framework embeds dataset geometry into MLP weights
Researchers have developed S-GAI, a novel initialization framework for sigmoidal MLPs that embeds dataset geometry directly into network weights. This method uses singular value decomposition (SVD) to estimate class-wis…
-
Full-resolution MLPs outperform CNNs and transformers in medical dense prediction
Researchers have developed a new framework for medical dense prediction tasks that utilizes Multi-layer Perceptrons (MLPs) at full image resolution. This approach aims to overcome limitations of Convolutional Neural Net…
-
GS-KAN offers parameter-efficient alternative to Kolmogorov-Arnold Networks
Researchers have introduced GS-KAN, a novel architecture that enhances the efficiency of Kolmogorov-Arnold Networks (KANs). By utilizing shared basis functions and learnable linear transformations, GS-KAN significantly …
-
New PURe Networks Explicitly Model Nonlinear Feature Interactions
Researchers have introduced Product-Unit Residual Networks (PURe) to better model nonlinear feature interactions in scientific and engineering applications. These networks integrate multiplicative product units with res…
-
Neural implicit learning reconstructs 3D geometry from RF signals
Researchers have developed GeRaF, a novel method for 3D geometry reconstruction using radio frequency (RF) signals. This approach leverages neural implicit learning to overcome the challenges of low resolution and noise…
-
New Ansatz Predicts Bayesian Deep Neural Network Performance
Researchers have developed a new approximate method to predict the generalization performance of Bayesian deep neural networks (MLPs) with fixed depth. The approach utilizes an equivalent Wishart Ansatz to model the flu…
-
Local SGD Worker Disagreement Reveals Deep Neural Network Loss Geometry
Researchers have developed a novel method to understand the loss geometry of deep neural networks by analyzing worker disagreement in Local Stochastic Gradient Descent (SGD). This disagreement, theoretically shown to be…
-
New KAN variants tackle efficiency and hardware implementation
Researchers have developed a new variant of Kolmogorov-Arnold Networks (KANs) called Kolmogorov-Arnold Fourier Networks (KAFs) to address limitations in parameter efficiency and high-frequency feature capture. KAFs repa…
-
New research explores neuron specialization in MLPs for data reconstruction
Researchers have investigated how training biases in minimal MLPs can lead to neuron specialization and improve the reconstruction of training data from learned weights. Experiments on one-dimensional datasets demonstra…
-
New tensor algebra embeds equivariance for symmetry discovery
Researchers have developed a new tensor algebra framework called $\star_G$ that intrinsically embeds equivariance, allowing for symmetry-preserving tensor approximation and physical symmetry discovery. This framework of…
-
New theory shows front-loaded dropout cuts neural network test loss
Researchers have developed a new mean-field theory for dropout, a technique used in neural networks. This theory suggests that by scheduling dropout to be more aggressive at the beginning of training, test loss can be r…
-
New theory explains dropout universality in neural networks
Researchers have developed a mean-field theory to understand dropout in neural networks, viewing it as a perturbation of critical signal propagation. The theory establishes distinct universality classes for smooth and R…
-
Paper questions weight decay's role in deep learning stability
A new paper investigates the role of weight decay in deep learning training stability, challenging its common perception as a simple regularization technique. The research analyzes how weight decay affects parameter dyn…
-
TabPFN model advances clinical decision support for pediatric ECMO
Researchers have developed an imitation learning approach to aid clinical decision-making for pediatric ECMO patients. This method uses observational data to learn action models, addressing challenges of complexity and …
-
KANs gain temporal explanations with new Temporal Functional Circuits
Researchers have developed a new framework called Temporal Functional Circuits to enhance the interpretability of Kolmogorov-Arnold Networks (KANs) in time-series forecasting. This method transforms the KAN's internal e…
-
KANs enable ultrafast on-chip online learning for low-latency systems
Researchers have demonstrated ultrafast online learning capabilities using Kolmogorov-Arnold Networks (KANs) on Field-Programmable Gate Arrays (FPGAs). This approach achieves sub-microsecond adaptation times, outperform…
-
Singular Bayesian Neural Networks
Researchers have introduced Singular Bayesian Neural Networks, a novel approach that significantly reduces the parameter count required for Bayesian neural networks. By parameterizing weights using a low-rank decomposit…