arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2103.05787cs.LG
收藏

基于柱状神经网络的可扩展在线递归学习

Scalable Online Recurrent Learning Using Columnar Neural Networks

University of Alberta(阿尔伯塔大学);DeepMind(深度思维公司)

Khurram Javed, Martha White, Rich Sutton

摘要:

结构性信用分配的循环学习是具有挑战性的。一种名为RTRL的算法可以在线计算递归网络的梯度,但对于大型网络来说,计算难度很大。替代方案,如BPTT,不在网上。在这项工作中,我们提出了一个信用分配算法- \algoname{} -近似梯度的循环学习实时使用$O(n)$操作和内存每一步。我们的方法建立在这样的思想基础上,即对于由具有标量状态的列组成的模块化递归网络,参数只跟踪其对列状态的影响就足够了。我们的经验表明,只要列之间的连接是稀疏的,我们的方法近似的真实梯度。在列之间没有连接的特殊情况下,$O(n)$梯度估计是准确的。我们证明了实用的方法,经常性的状态学习和元学习的估计梯度的真实梯度的合成测试床进行比较。

英文摘要:

Structural credit assignment for recurrent learning is challenging. An algorithm called RTRL can compute gradients for recurrent networks online but is computationally intractable for large networks. Alternatives, such as BPTT, are not online. In this work, we propose a credit-assignment algorithm -- \algoname{} -- that approximates the gradients for recurrent learning in real-time using $O(n)$ operations and memory per-step. Our method builds on the idea that for modular recurrent networks, composed of columns with scalar states, it is sufficient for a parameter to only track its influence on the state of its column. We empirically show that as long as connections between columns are sparse, our method approximates the true gradient well. In the special case when there are no connections between columns, the $O(n)$ gradient estimate is exact. We demonstrate the utility of the approach for both recurrent state learning and meta-learning by comparing the estimated gradient to the true gradient on a synthetic test-bed.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑