Neural Networks news and analysis on Web Pulse: 747 curated articles tracking the latest Neural Networks developments, tools, and research, updated continuously from vetted sources.
A team from Tsinghua University, the Chinese Academy of Sciences, and the University of Tokyo developed an all-organic, flexible brain-computer interface chip that retained 94% signal efficiency after 18 months in animal…
A loss function measures how wrong a machine learning model's predictions are, providing a feedback signal for improvement. High loss indicates large errors, while low loss means the model is close to correct. Common los…
A developer implemented real-time LiDAR room detection on an Arduino UNO Q using Edge Impulse, achieving high-accuracy environment classification without SLAM or computer vision. The system uses a 2D dToF LiDAR to create…
Sungkyunkwan University researchers led by professors Jo Sae-byeok and Yang Woo-seok developed a semiconductor synapse that strengthens or weakens memory by changing light wavelength, using disorder in an AgBiS2 heterost…
Sungkyunkwan University researchers developed a technology that strengthens or weakens AI semiconductor memory by changing light color, using material disorder and defects to mimic brain memory homeostasis. The team expe…
Flourish Labs has secured $500 million in funding to develop a new approach to artificial intelligence based on neuroscience principles. The company aims to reinvent AI by leveraging insights from brain function, potenti…
Researchers introduced the Integral Transform Network (ITNet), a unified architecture that subsumes convolution, attention, and recurrence as special cases of a learnable integral transform. ITNet matches or exceeds spec…
A new study measures the linearity of transformer feed-forward blocks, finding that linear recoverability (R²_lin) varies widely across blocks and is a learned property, not an architectural one. The analysis of GPT-2, P…
Researchers proposed a hybrid GNN-FEM framework for phase-field fracture simulation that integrates a graph neural network surrogate into the conventional staggered scheme, replacing the phase-field update while retainin…
Researchers propose a physics-informed framework to discover anisotropic yield functions from full-field displacement and reaction force data, using a convex neural network representation. The method avoids direct stress…
Researchers introduced SSProNet, a secondary-structure-aware graph neural network for protein representation learning that uses energy-filtered hydrogen-bond graphs. The model outperforms existing graph-based methods on …
Researchers introduced LooseControlVideo, a framework that uses sparse, oriented 3D boxes as a proxy for spatial blocking in text-to-video generation, enabling intuitive control over layout and trajectory. Fine-tuned on …
Researchers propose LEAP, a training curriculum for Vision Transformer feature-based knowledge distillation that uses the teacher's intermediate feature maps as progressively difficult targets. LEAP achieves 90.1% accura…
Researchers derived a three-force decomposition of AdamW training dynamics explaining Weibull weight-scale parameter evolution in transformers, finding alignment force dominates growth. A spline displacement method recov…
Integer quantization reduces neural network memory and energy costs by converting high-precision values to lower-bit integers, with INT8 additions consuming 30 times less energy than FP32. The technique leverages hardwar…
Chinese scientists developed a brain implant electrode array thinner than a hair and softer than brain tissue, which recorded neural activity with high clarity for 18 months in animal trials without performance decline, …
A new research piece from Rudrite Research compares three generative modeling approaches: Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Diffusion Models, explaining how each learns a data d…
The Continuum Memory System introduces a novel architecture that replaces the Transformer's two-rate memory with a spectrum of blocks, each updating on its own clock, potentially improving efficiency and performance in A…
A developer explains that self-attention is a matrix operation, not just a mechanism for looking at important words. The post breaks down the QKV projection, scaling, softmax, and weighted sum that enable Transformers to…
Scientists at Harvard Medical School discovered that the stress hormone cortisol helps lock in early neural connections by signaling astrocytes to close critical periods of brain plasticity. The study, published in Natur…