› models
ML/AI Writing
Machine learning notes, model architecture explainers, and practical AI write-ups.
Trust Region Policy Optimization, Proximal Policy Optimization
How TRPO and PPO keep policy updates stable in reinforcement learning, and where off-policy methods reuse past experience to get there.
GPT, PaLM, GLM, LLaMA and beyond: Large language models in medicine
GPT, PaLM, GLM, and LLaMA applied to medicine: where they are already useful, where they fall short, and what the open challenges are.
Power of Multimodal Machine Learning: Challenges, Innovations, and Future Directions
What makes multimodal machine learning hard, which recent innovations moved it forward, and where integration across modalities is heading.
Understanding Key Insights and Trends in Large Language Models
How large language models are built and how they have evolved, traced through the key insights and trends behind each generation.
What is Deep Reinforcement Learning (RL)
The language and notation of reinforcement learning, and what an RL agent actually does, introduced from first principles.
An in-Depth Look at Leading Computer Vision ML Models
The leading computer vision model families in one place, from the architecture each uses to the task it is actually good at.
What is OpenMMLab and how it is used in computer vision
OpenMMLab as a computer vision platform: what MMDetection, MMSegmentation, MMPose, MMOCR, and MMDeploy each cover, and where they fit in a workflow.
YOLO and object detection
How the YOLO object detection series changed from v1 onward, and what each version actually improved over the one before it.
Yet another gentle intro to the YOLO-NAS model
How Deci AI's YOLO-NAS combines neural architecture search with YOLO object detection, from its multi-phase training to reading the model's output.
Fooocus Fantasia: Navigating Generative Magic
A working guide to Fooocus, the image generation interface that hides the tuning, and how to run it locally or on Colab.
Behind Generative Models [In Progress]
The mechanics behind diffusion models, and what separates DALL-E, Midjourney, and Stable Diffusion once you look past the interface.
The Generative Lens: Crafting Camera Angles with Precision
A guide to refining camera angle outputs in generative platforms using nuanced prompt strategies.
Q-learning
Q-learning is a reinforcement learning technique where an agent iteratively learns the value of its actions to navigate towards optimal outcomes in an environment.