edge_ai Computer Science

3 个科学问题 · 计算机 Computer Science

本页目录

模型剪枝(pruning)是如何压缩神经网络的?

How does model pruning compress a neural network?

剪枝会识别并移除权重中接近零或贡献很小的连接(或整个通道),把稠密网络变稀疏,从而减少参数量和计算量;之后常配合稀疏算子或再训练来恢复精度。
Pruning removes connections (or whole channels) with near-zero or low contribution, turning a dense network sparse to cut parameters and compute; it is often followed by sparse kernels or retraining to recover accuracy.
难度 专业 · 适合 大众 CC BY 4.0 · 已核验 · Verified 2026-08-23 · 2 条独立证据
来源 · Source

Russell & Norvig,《Artificial Intelligence: A Modern Approach》第4版,Pearson,2020

S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed., Pearson, 2020

#剪枝 #压缩

边缘 AI 为什么强调在设备本地做推理?

Why does edge AI emphasize running inference on the device itself?

在手机、摄像头等终端本地推理可避免把数据传上云,从而降低延迟、节省带宽、保护隐私,并在断网时仍可用;代价是终端算力与电量有限,需要更小的模型。
Running inference on phones or cameras avoids sending data to the cloud, cutting latency, saving bandwidth, protecting privacy, and working offline—at the cost of limited compute and battery, needing smaller models.
难度 进阶 · 适合 大众 CC BY 4.0 · 已核验 · Verified 2026-08-23 · 2 条独立证据
来源 · Source

Russell & Norvig,《Artificial Intelligence: A Modern Approach》第4版,Pearson,2020

S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed., Pearson, 2020

#边缘 #本地

为什么现在的手机也能跑人工智能?

Why can today's phones run AI?

因为芯片里加入了专为矩阵运算设计的 NPU,并且模型被压缩得很小,再加上很多计算在手机本地完成,所以语音助手、拍照美颜等 AI 功能能在不打云的情况下运行。
Phones now have NPUs built for matrix math, models are shrunk small, and much computing happens on the device, so features like voice assistants and camera AI run without the cloud.
难度 入门 · 适合 大众 CC BY 4.0 · 已核验 · Verified 2026-08-23 · 2 条独立证据
来源 · Source

Russell & Norvig,《Artificial Intelligence: A Modern Approach》第4版,Pearson,2020

S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed., Pearson, 2020

#手机 #端侧