تقرير
WrapNet: Neural Net Inference with Ultra-Low-Resolution Arithmetic
العنوان: | WrapNet: Neural Net Inference with Ultra-Low-Resolution Arithmetic |
---|---|
المؤلفون: | Ni, Renkun, Chu, Hong-min, Castañeda, Oscar, Chiang, Ping-yeh, Studer, Christoph, Goldstein, Tom |
سنة النشر: | 2020 |
المجموعة: | Computer Science Statistics |
مصطلحات موضوعية: | Computer Science - Machine Learning, Statistics - Machine Learning |
الوصف: | Low-resolution neural networks represent both weights and activations with few bits, drastically reducing the multiplication complexity. Nonetheless, these products are accumulated using high-resolution (typically 32-bit) additions, an operation that dominates the arithmetic complexity of inference when using extreme quantization (e.g., binary weights). To further optimize inference, we propose a method that adapts neural networks to use low-resolution (8-bit) additions in the accumulators, achieving classification accuracy comparable to their 32-bit counterparts. We achieve resilience to low-resolution accumulation by inserting a cyclic activation layer, as well as an overflow penalty regularizer. We demonstrate the efficacy of our approach on both software and hardware platforms. |
نوع الوثيقة: | Working Paper |
URL الوصول: | http://arxiv.org/abs/2007.13242 |
رقم الأكسشن: | edsarx.2007.13242 |
قاعدة البيانات: | arXiv |
الوصف غير متاح. |