Neural Networks news and analysis on Web Pulse: 1652 curated articles tracking the latest Neural Networks developments, tools, and research, updated continuously from vetted sources.
A new method for neural network compression learns functional subspaces to select which low-rank components to remove from each transformer weight matrix, according to the paper "Learning Functional Subspaces for Neural …
Researchers at the Weizmann Institute of Science, led by Michal Irani, developed an AI "brain decoder" that reconstructs images a person is viewing from fMRI brain scans, training on a public dataset of scans from eight …
Scientists developed an AI model that reconstructs images from a person's brain activity at the fastest speed and highest accuracy reported to date, according to the researchers behind the work. The model reads brain act…
A physics-informed neural network (PINN) was benchmarked against a plain finite-difference solver in two experiments, one in 1D and one in 5D, with the winning method changing between the two dimensions, according to a T…
A new technical note explains that in a shallow homogeneous neural net, minimizing the L2 norm of the parameters is equivalent to minimizing the L1 norm of a measure over unit-norm neurons, a duality the author says is d…
AbstractPhil released Beatrix V3, a 376-million-parameter byte-level language model with a 256-token vocabulary and 4,096-byte context, trained on 64.4 billion bytes across two RTX 5090 cards in 295 hours to 0.9861 bits …
A graph neural network combining facial and upper-body skeletal keypoints classified simulated train-driver states into alert, not-alert and emergency classes with 81% accuracy in light conditions, according to a study p…
Researchers introduced TEMPEST, a Temporal Convolutional Network embedding model trained with an additive angular margin (ArcFace) loss that maps 60-second multimodal driving windows to 96-dimensional embeddings for scal…
Researchers introduced rollout-verified policy improvement (RV-PI), a long-horizon reinforcement learning method for budgeted neural-operator PDE solvers, detailed in arXiv paper 2610.06883v1. On the shallow-water benchm…
Researchers posted arXiv:2610.06942v1, proposing Memory-Discrepancy Knowledge Distillation (MemKD), a knowledge distillation framework that transfers memory retention behavior from a large teacher recurrent neural networ…
A new arXiv paper (2610.06952v1) proposes an evaluation framework showing that neural PDE solvers with low prediction errors can still fail to reproduce the underlying dynamics of the systems they model. Testing on two-d…
A new arXiv thesis (arXiv:2610.06974v1) systematically investigates three sources of uncertainty in knowledge graph embedding (KGE): knowledge uncertainty from incomplete, noisy, or probabilistic input knowledge; algorit…
A Vision Transformer-based model with dual-pathway temporal and spatial attention achieved the highest deterministic and probabilistic skill at all lead times for subseasonal soil-moisture forecasting over Europe, outper…
CNet, a C++/CUDA framework for building and training complex-valued neural networks, was released on arXiv as paper 2610.08592v1 with code at github.com/crasmarum/CNet. The framework implements Wirtinger (CR-calculus) au…
Researchers introduced BeFOND, an encoder-free sparse coding model that unifies inference and dictionary learning as natural-gradient flows on a shared variational free energy, according to arXiv paper 2610.07389v1. On s…
Researchers introduced learnable spectral activations (LSA), a method that replaces fixed neuron-level nonlinearities with a residual truncated Fourier series whose harmonic amplitudes are learned during training, accord…
A York University study led by neuroscientist Kohitij Kar found that nine AI vision networks failed to reproduce the human "motion aftereffect" illusion, while two macaques tested alongside 79 humans showed responses sim…
Researchers at the Weizmann Institute of Science developed Brain-IT, an AI system that reconstructs viewed images from fMRI brain scans, training on more than 70,000 images collected from eight subjects and identifying 1…
A component-level decomposition study of Dr. Bojan Ploj's bipropagation method finds that per-layer supervision, not gradient locality, is the ingredient that makes greedy layer-wise supervised training effective, with a…
Researchers introduced HLA-WM, a hybrid linear attention architecture for long-horizon video world models that combines softmax attention's full-history KV cache with recurrent linear attention's fixed-size state compres…