Explaining Convolution in Simple Terms

Explains the meaning of convolution—why we flip (fold) and multiply—using signal analysis, dice probability, and image processing kernels as examples.

PixelLM: First Efficient Pixel-Level Inference Model Without SAM

PixelLM is an efficient pixel-level multimodal large model for open-domain multi-object reasoning segmentation without SAM, accompanied by the MUSE dataset.

Future Directions in Lightweight Deep Learning

Review of lightweight deep learning for resource-constrained devices: TinyML, quantization, architectures and deployment strategies for efficient inference.

Practical Tricks for Reward Models

Technical overview of methods to improve reward model robustness for RLHF: quantify preference strength, flip/soften labels, apply adaptive margins, contrastive learning and MetaRM

iTOP-RK3568 MobileNet Image Classification Inference

Guide to validating MobileNet image classification inference on the iTOP-RK3568 board, covering RK3568 hardware, NPU use, model files, and execution steps.

Passive Component Challenges and Solutions in AI

Overview of passive components for AI systems: material, architectural and process innovations for high-current power inductors and low-ESR polymer tantalum capacitors.

Can LLMs Reason and Plan?

Analysis of recent research evaluating whether LLMs can plan or reason, showing limited autonomous planning and that apparent emergent capabilities stem from in-context learning.

Crystal Oscillators in AI Smart Glasses

Overview of AI glasses hardware and the role of quartz crystal oscillators in providing precise clock signals for display rendering, data processing, and wireless stabilization.

AI Applications for Improving Industrial Automation Stability

Predictive maintenance and digital twin applications for AI-driven industrial process stability, enabling real-time parameter optimization and reduced downtime.

AIoT Connecting AI to the Physical World

Explains how AIoT links AI and physical devices via IP-based networks and application layers like Matter, and deployment considerations for scalable device connectivity.