arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2312.03802physics.optics
收藏

集成光子神经网络中的涌现式自适应实现无反向传播学习

Emergent self-adaptation in an integrated photonic neural network for backpropagation-free learning

Alessio Lugnan, Samarth Aggarwal, Frank Brückerhoff-Plückelmann, C. David Wright, Wolfram H. P. Pernice, Harish Bhaskaran, Peter Bienstman

AI总结:

本文提出一种由硅光子微环谐振器与相变材料构成的光子神经网络,通过涌现式自适应机制实现无反向传播学习,在MNIST手写数字分类任务上达到98.2%的准确率,为高速实现生物合理性学习方案提供了独特平台。

摘要:

可塑性自适应、非线性循环动力学和多尺度记忆是神经网络硬件实现中备受期待的特性,因为它们使硬件能够以类似于生物大脑的方式学习、适应和处理信息。在这项工作中,我们通过实验证明了这些特性在光子神经元阵列中的出现。重要的是,这一过程以涌现的方式自主实现,无需外部控制器设置权重,也无需全局奖励信号的显式反馈。通过将此类阵列的层级结构与基于简单逻辑回归的无反向传播训练算法相结合,我们在MNIST任务上取得了98.2%的性能,该任务是一个关注手写数字分类的常用基准任务。可塑性节点由硅光子微环谐振器构成,其上覆盖有一片实现非易失性存储的相变材料。该系统结构紧凑、鲁棒性强,并且通过使用多个波长可以方便地扩展规模。此外,它构成了一个独特的平台,能够在高处理速度下测试并高效实现生物学上合理的学习方案。

英文摘要:

Plastic self-adaptation, nonlinear recurrent dynamics and multi-scale memory are desired features in hardware implementations of neural networks, because they enable them to learn, adapt and process information similarly to the way biological brains do. In this work, we experimentally demonstrate these properties occurring in arrays of photonic neurons. Importantly, this is realised autonomously in an emergent fashion, without the need for an external controller setting weights and without explicit feedback of a global reward signal. Using a hierarchy of such arrays coupled to a backpropagation-free training algorithm based on simple logistic regression, we are able to achieve a performance of 98.2% on the MNIST task, a popular benchmark task looking at classification of written digits. The plastic nodes consist of silicon photonics microring resonators covered by a patch of phase-change material that implements nonvolatile memory. The system is compact, robust, and straightforward to scale up through the use of multiple wavelengths. Moreover, it constitutes a unique platform to test and efficiently implement biologically plausible learning schemes at a high processing speed.

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。

↑