Yet another disentangled VAE ... but for quick drawing doodles

🛰️ Explore NASA data with an all-in-one app featuring APOD, image search, live Earth events, and Near-Earth Objects tracking.

World Structured SwiGLU FFN

Chem-MRL: SMILES-based Matryoshka Representation Learning Embedding Model

Prescriptive representation engineering experiments

Free Transformer Architecture - an extension of the decoder Transformer that conditions its generative process on random latent variables which are learned without supervision

Official implementation of the paper "PonderLM-2: Pretraining LLM with Latent Thoughts in Continuous Space"

Deep learning pipeline to obtain latent representation from images using 2D and 3D diffusion and other autoencoders

Deterministic, offline-first agent sandbox with concept-graph memory. Reflex→retrieval→planning→filter for repeatable outputs. Stable baseline for v4 external reasoning.

Stable Diffusion

Create 3DGS heads from latent vectors

ILVE is an interactive Streamlit framework for exploring and understanding VAE/β-VAE latent spaces through real-time visualization, systematic β-parameter experiments, and educational tools for deep learning research.

Medical image classification of skin lesions using PyTorch and scikit-learn. Combines a Deep Learning pipeline (Convolutional Autoencoder + neural classifier) with a hybrid approach leveraging Vision Transformer embeddings and XGBoost for robust and efficient model comparison.

Hướng dẫn và mã nguồn về Suy luận Biến thiên và Mạng Nơ-ron (Autoencoder & VAE) cho bài toán xấp xỉ, trực quan hóa không gian ẩn với các thí nghiệm trên không gian 2D và 32D.

Synthetic Data Generation for Emotional Depth Faces: Optimizing Conditional DCGANs via Genetic Algorithms in the Latent Space and Stabilizing Training with Knowledge Distillation

Pytorch + HuggingFace Diffusers project that experiments with latent diffusion model by training variational autoencoder (VAE) and a conditional U-Net.

Open-source AI system that uses artificial gravity fields to compress and fuse multimodal data with sub-ms inference. CPU-only, 5k param model. Demo of a novel architecture built for reasoning, not token prediction.

A deep learning architecture for robust inference and accurate prediction of cellular dynamics

Implementation of 🥥 Coconut, Chain of Continuous Thought, in Pytorch

[ICASSP 2025 Oral] ImageFlowNet: Forecasting Multiscale Image-Level Trajectories of Disease Progression with Irregularly-Sampled Longitudinal Medical Images