The neo labs don’t want you to know about convolutional interpretability for language models. They are suppressing it i can’t say much more without getting into tr—

💥Excited for the publication: "A Hybrid Convolutional–Transformer Approach for Accurate Electroencephalography (EEG)-Based Parkinson’s Disease Detection" 🔗brnw.ch/21x3jlM 📌 #Parkinsons #EEG #ArtificialIntelligence #Neuroscience #BiomedicalEngineering

#20 Coding Convolutional Neural Network (CNN) - Teknik Informatika

#5 zu Architekturen wie Convolutional Neural Networks (CNN) mit konkreten Ausprägungen wie das klassische U-Net und Transformer-basierten Architekturen Vision Transformer (ViT) und Diffusion Transformer (DiT) mit Fokus auf die Konzepte und weniger die exakte Mathematik. 1/10

It’s not about pouring $50–60 billion into AI. What we really need are exceptional scientists in physics, mathematics, and computer science,researchers of the same caliber as the godfathers of deep learning: Geoffrey Hinton, Yann LeCun, and Yoshua Bengio.Geoffrey Hinton, Widely regarded as the “Godfather of Deep Learning.” His foundational work on backpropagation, Boltzmann machines, and deep belief networks helped shape modern AI. He shared the 2024 Nobel Prize in Physics with John Hopfield. Yann LeCun — Pioneered convolutional neural networks (CNNs), the cornerstone of modern computer vision, and has long led AI research at Meta. Yoshua Bengio — Made critical advances in recurrent networks, probabilistic models, and the theoretical foundations of deep learning. These three researchers are collectively known as the “Godfathers of Deep Learning” for their transformative contributions to the field.

A deep convolutional neural network model predicts the influence of the full-length mRNA sequence on translation efficiency go.nature.com/4mgXR0C rdcu.be/flWMe

spaMGCN: a graph convolutional network with autoencoder for spatial domain identification using multi-scale adaptation genomebiology.biomedcentral.…

Language convolutional autoencoders, enhancing...

On sequencing depth robustness: Velocyto held up best at low read counts, followed by scv-Sto. DeepVelo was the most sensitive to downsampling — its graph convolutional architecture shifts when the data changes. Something to weigh before picking a method for shallow datasets.

I just published ImageNet Classification with Deep Convolutional Neural Networks (AlexNet) medium.com/p/imagenet-classi…

"We’ve seen a massive shift from convolutional networks to the new Transformer architectures that power today’s [LLMs], but the way these networks route information from one layer to another hasn’t changed all that much." @SteadySurdom explores new research that aims to address this status quo. towardsdatascience.com/why-t…

What happens when you put a convolutional block in the middle of a language model and then look at the activations as it's generating?