📄 中文摘要
彩票假设是由Jonathan Frankle和Michael Carbin于2018年首次提出的一个革命性概念，揭示了在人工智能领域中，神经网络训练的潜在高效性。该假设将大型神经网络比作一组彩票，其中大多数彩票是无用的，但存在一些稀疏子网络，如果能够被识别并单独训练，能够达到与完整网络相当的性能。这一发现为神经网络的优化和资源利用提供了新的视角，可能在未来的深度学习研究中发挥重要作用。
彩票假设：一种可以高效训练的稀疏神经网络发现
📄 English Summary
Hipotesis Tiket Lotre: Penemuan Jaringan Saraf Jarang yang Dapat Dilatih dengan Efisien
The Lottery Ticket Hypothesis, introduced by Jonathan Frankle and Michael Carbin in 2018, reveals the potential for extraordinary efficiency in training artificial neural networks. This hypothesis analogizes large neural networks to a collection of lottery tickets, where most tickets are useless, but there exist sparse subnetworks that, if identified and trained separately, can achieve performance comparable to the full network. This finding offers a new perspective on optimizing neural networks and resource utilization, potentially playing a significant role in future deep learning research.