使用混乱光的概率光子计算
Frank Brückerhoff-Plückelmann1,2, Hendrik Borras3, Bernhard Klein3
1Physical Institute, University of Münster, Münster, 48149, Germany.
Nature communications
|December 1, 2024
概括
这项研究引入了一种用于超快速概率计算的新型光子方法. 它使人工神经网络能够量化预测不确定性,通过实时不确定性估计增强图像分类.
科学领域:
- 人工智能的人工智能
- 计算神经科学是一种神经科学.
- 光学和光子学 在光学和光子学.
背景情况:
- 生物神经网络擅长进行复杂的计算和处理杂数据.
- 人工神经网络 (ANN) 是强大的,但通常提供点估计,缺乏不确定性量化.
- 对ANN的贝叶斯推理对传统架构提出了计算挑战.
研究的目的:
- 开发使用混乱光和光子数据处理的高速概率计算架构.
- 为了使ANN能够同时进行图像分类和不确定性预测.
- 将物理源与用于超快速概率计算的计算架构集成.
主要方法:
- 利用混乱的光和不连贯的光子数据处理进行概率计算.
- 在光子架构中实现贝叶斯神经网络.
- 采用并行采样用于高速计算和不确定性量化.
主要成果:
- 使用光子概率架构展示了同时的图像分类和不确定性预测.
- 通过并行采样实现了高速概率计算.
- 成功地将物理源与计算框架集成.
结论:
- 光子概率计算为克服传统ANN在不确定性量化方面的局限性提供了一条途径.
- 这种方法可以对复杂的数据进行超快速,不确定性意识的预测.
- 展示的原型为需要强大的不确定性估计的先进AI应用铺平了道路.
相关概念视频
02:12The Wave Nature of Light
48.4K
The nature of light has been a subject of inquiry since antiquity. In the seventeenth century, Isaac Newton performed experiments with lenses and prisms and was able to demonstrate that white light consists of the individual colors of the rainbow combined together. Newton explained his optics findings in terms of a "corpuscular" view of light, in which light was composed of streams of extremely tiny particles traveling at high speeds according to Newton's laws of motion.
48.4K
02:32The de Broglie Wavelength
25.3K
In the macroscopic world, objects that are large enough to be seen by the naked eye follow the rules of classical physics. A billiard ball moving on a table will behave like a particle; it will continue traveling in a straight line unless it collides with another ball, or it is acted on by some other force, such as friction. The ball has a well-defined position and velocity or well-defined momentum, p = mv, which is defined by mass m and velocity v at any given moment. This is the typical...
25.3K
01:10Entropy Change in Reversible Processes
2.5K
In the Carnot engine, which achieves the maximum efficiency between two reservoirs of fixed temperatures, the total change in entropy is zero. The observation can be generalized by considering any reversible cyclic process consisting of many Carnot cycles. Thus, it can be stated that the total entropy change of any ideal reversible cycle is zero.
The statement can be further generalized to prove that entropy is a state function. Take a cyclic process between any two points on a p-V diagram.
The statement can be further generalized to prove that entropy is a state function. Take a cyclic process between any two points on a p-V diagram.
2.5K