New HiRo Model Achieves High Accuracy with Under 1M Parameters

Researchers have introduced HiRo, a novel image classification model designed for efficiency and performance. HiRo utilizes a combination of shifted-window partitioning and multi-directional hierarchical reservoir computing to process image tokens. This approach allows for effective local feature modeling and cross-window interaction while maintaining a low parameter count, under 1 million trainable parameters. The model demonstrates strong accuracy on benchmark datasets, achieving 99.46% on MNIST, 85.57% on CIFAR-10, and 59.10% on CIFAR-100. AI

IMPACT Introduces a parameter-efficient architecture for image classification, potentially reducing computational costs for AI systems.

RANK_REASON The cluster describes a new academic paper detailing a novel model architecture and its performance on benchmark datasets. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• CIFAR-10
• CIFAR-100
• MNIST database

• paper
• model release
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →