Binary Neural Networks

Also known as BNN

Overview

Binary neural networks represent weights, activations, or both with a single bit, aiming to cut memory use and replace costly multiplications with simpler operations. Training methods include straight-through estimators, smooth or stochastic relaxations, progressive binarization, and auxiliary full-precision paths to address the difficulty of optimizing discrete sign functions. Hardware work maps these models to FPGAs, resistive-memory crossbars, and other low-power devices, while applications include vision, event processing, continual learning, and compact language models. The main tension is retaining accuracy and expressive capacity as precision falls; work also examines how training stability, robustness, and generalization depend on architecture and task.

Latest papers

CardsList

33 papers4 papers in the last four weeks, against 2 the four weeks before. 0.0% of all new papers.

  1. Highly Efficient and Effective LLMs with Multi-Boolean Architectures

    May 28, 2025Ba-Hien Tran, Van Minh NguyenLLM QuantizationBinary Neural Networks

  2. Optimizing Binary and Ternary Neural Network Inference on RRAM Crossbars using CIM-Explorer

    May 20, 2025Rebecca Pelke, José Cubero-Cascante, Nils Bosbach +5Efficient Neural Network InferenceCompute-in-Memory

  3. Stochastic Engrams for Efficient Continual Learning

    Mar 27, 2025Isabelle Aguilar, Luis Fernando Herbozo Contreras, Omid KaveheiClass-Incremental LearningBinary Neural Networks