residual neural network
PulseAugur coverage of residual neural network — every cluster mentioning residual neural network across labs, papers, and developer communities, ranked by signal.
- instance of ResNet-18 95%
- instance of alphaXiv 90%
- instance of CatalyzeX 90%
- instance of DagsHub 90%
- instance of Gotit.pub 90%
- instance of VideoGameGeek 90%
- instance of Deep Neural Networks 90%
- instance of ResNet-34 90%
- used by alphaXiv 70%
- instance of ScienceCast 70%
- used by Gotit.pub 70%
- used by ScienceCast 70%
8 day(s) with sentiment data
-
New methods enhance vision-language model adaptation with limited supervision · 3 sources tracked
Researchers are developing new methods for adapting vision-language models (VLMs) to new domains with limited supervision. One approach, ACTOR, uses a training-free plug-in module that leverages queried labels from futu…
-
AI model predicts stroke outcomes using MRI and clinical data
Researchers have developed a method to predict functional outcomes after acute ischemic stroke (AIS) by fusing multimodal MRI data with clinical information. The study found that combining FLAIR MRI with clinical data y…
-
New CLS model enhances few-shot and continual learning by separating memory systems
Researchers have developed a new model inspired by the complementary learning system (CLS) architecture found in the hippocampus. This model aims to balance generalization with continuous adaptation by separating short-…
-
ConvNeXt Explained: CNNs Adopt Vision Transformer Techniques
This article delves into the ConvNeXt architecture, explaining how Convolutional Neural Networks (CNNs) have incorporated elements from Vision Transformers. It explores the design choices made in ConvNeXt and contrasts …
-
New ensemble head boosts vision model efficiency for plant stage classification
Researchers have developed a novel multibranch ensemble head technique for vision foundation models, aiming to improve efficiency and accuracy when adapting these models to smaller, domain-specific datasets. This method…
-
AI models offer switchable unlearning and costly reasoning
A new framework called MoME allows retrieval models to selectively forget specific individuals while retaining general recognition capabilities. This is achieved by splitting the model into core parameters for retrieval…
-
New Control Theory-Based Data Augmentation Algorithm Unveiled
Researchers have introduced ControlAugment (Ctrl-A), a novel data augmentation algorithm for image-vision tasks that leverages control theory principles. This method dynamically adjusts augmentation strength distributio…
-
New research questions foundational premise of neural network pruning
A new paper questions the long-standing foundation of neural network pruning, known as oracle pruning. The research, which involved training and analyzing 37,000 models across various architectures and datasets, found t…
-
ResNet research explained: Deeper networks, shortcut solutions
This article breaks down the research behind residual neural networks (ResNets), explaining why deeper neural networks initially performed worse and how shortcut connections provided a solution. It highlights the develo…
-
Hybrid Mamba-Transformer Models Advance 3D Reconstruction and Brain Aging Analysis
Researchers have developed two novel hybrid deep learning models that combine Mamba and Transformer architectures for distinct applications. The first, a Hybrid Mamba-Transformer Network, is designed for single-shot 3D …
-
New research explores functional modularity in neural networks and efficient SNN quantization
Two new research papers explore advancements in neural network architectures and efficiency. The first paper, "Sparsity-driven emergence of functional modularity in multitask neural networks," demonstrates how sparsity-…
-
New method extracts generative capabilities from discriminative AI models
Researchers have developed a new method called the Discriminative Score Function (DSF) that can extract generative capabilities from pretrained discriminative models. This technique allows models trained for tasks like …
-
何恺明团队发布VISTA视觉Harness，提升多模态模型推理能力 · 1 source tracked
Researchers led by何恺明 (Kaiming He) have introduced VISTA, a novel visual harness designed to enhance multimodal models' ability to reason within interactive environments. VISTA allows models to retain and revisit raw vi…
-
Naive AI raises $400M in 7 months with elite Tsinghua-led research team
Naive AI, a startup founded by Tsinghua University associate professor Dai Jifeng, has secured $400 million in funding within seven months, achieving a post-investment valuation of $1.42 billion. The company's core team…
-
Global Exponential Convergence Proven for Two-Layer Linear Networks
Researchers have proven global exponential convergence for training wide two-layer linear networks using smooth Polyak-Lojasiewicz predictor losses. The study demonstrates that gradient flow in the factors can be precis…
-
Deep learning model TTNet analyzes table tennis player data
Researchers have developed TTNet, a deep learning model designed for analyzing table tennis player data collected from smart rackets. This model utilizes a combination of convolutional neural networks, residual networks…
-
Stockfish chess engine value function distilled into neural network
A project has successfully distilled the Stockfish chess engine's value function into a neural network model, utilizing a ResNet and Vision Transformer architecture. This distilled model was trained on a massive dataset…
-
ResNets implicitly route by varying residual interaction experts
A new research paper explores the internal workings of Residual Networks (ResNets), suggesting they implicitly perform a form of soft routing. By analyzing ResNet-18 and ResNet-34 models trained on ImageNet, the study f…
-
Research paper links CNN topology to trainability, not just layer count
A new research paper explores the relationship between the nominal depth of convolutional neural networks (CNNs) and their trainability, introducing the concept of "effective depth." The study found that while nominal d…
-
New ANCRe framework optimizes neural network depth scaling
Researchers have introduced ANCRe, a novel framework designed to optimize the depth scaling of neural networks. By adaptively learning and reassigning residual connections, ANCRe aims to improve the utilization of deepe…