This book aims to demystify large deep networks and generative models, often perceived as "black boxes," by exploring their internal mechanisms through the lens of representation learning. It outlines how modern neural network architectures are designed, utilizing optimization and information theory.
This content refers to a research paper titled "Follow the Mean: Reference-Guided Flow Matching". It explores a new methodology in generative models.
![Follow the Mean: Reference-Guided Flow Matching [R]](/cdn-cgi/image/width=3840,quality=75,format=webp/https://preview.redd.it/5pleq5b4861h1.png?width=140&height=91&auto=webp&s=5f80ce290c30e51700f9b9fd0f907ee56e9382b2)
MOLPAQ is a novel modular quantum-classical generator that creates interpretable molecules from quantum-generated latent patches, achieving 100% RDKit validity and high novelty and diversity. This approach significantly improves property control, such as QED and aromatic motif incidence, over classical generators.
This research introduces Conditional Attribute Transformers, a novel method for jointly estimating next-token probability and an attribute's value conditional on each potential next token selection. This framework enables critical capabilities like per-token credit assignment and counterfactual analysis within a single forward pass, overcoming limitations of traditional generative models.
The paper introduces Dual-Spectral Flow Matching (DSFM), a novel fMRI generative framework that cascades dual frequency representation of BOLD signals with spectral flow matching. This method aims to synthesize high-fidelity fMRI time series to overcome limitations in data availability for brain disorder identification, by replicating complex spatiotemporal dynamics.
This research focuses on developing more efficient methods for sampling from Diffusion Probabilistic Models, aiming to reduce the computational cost and time associated with generating high-quality samples. It explores novel algorithms to accelerate the sampling process while maintaining the fidelity of the generated data.
Este artigo descreve um modelo generativo autorregressivo, treinado com dados de mais de 300.000 pacientes, capaz de simular trajetórias contrafactuais clinicamente plausíveis. O modelo foi aplicado a pacientes com COVID-19 para prever resultados com base em parâmetros clínicos alterados, demonstrando seu potencial para medicina personalizada e ensaios in silico.
UI-Oceanus é uma estrutura que escala agentes GUI generalistas, focando em dominar a física da interação através de feedback ambiental em vez de imitar trajetórias. O sistema utiliza exploração autônoma e predição de dinâmicas futuras para construir um modelo de mundo interno robusto, superando limitações de dados e supervisão.