Memory Capacity of Neural Networks with Threshold and Rectified Linear Unit Activations
Memory Capacity of Neural Networks with Threshold and Rectified Linear Unit Activations
复制标题
DOI：
10.1137/20m1314884
复制
发表时间：
2020-10
期刊：
影响因子：
--
通讯作者：
R. Vershynin
中科院分区：
文献类型：
--
作者：
R. Vershynin
压倒性的理论和经验证据表明，轻度过度参数化的神经网络--那些连接数量超过训练数据大小的神经网络--通常能够记住训练数据。
Overwhelming theoretical and empirical evidence shows that mildly overparametrized neural networks---those with more connections than the size of the training data---are often able to memorize the ...