没有处理器的机器学习:在非线性模拟网络中出现的学习
Sam Dillavou1, Benjamin D Beyer1, Menachem Stern1
1Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA 19104.
概括
研究人员开发了非线性电子对比本地学习网络 (CLLNs),以实现更快,更高效的模拟机器学习. 这种新的硬件可以完成复杂的任务,并显示出低功耗边缘计算的潜力.
科学领域:
- 电子工程 电子工程
- 机器学习硬件 机器学习硬件
- 模拟计算是一种模拟计算.
背景情况:
- 标准的深度学习需要对大型非线性网络进行缓慢,功率密集的差异化.
- 现有的电子对比本地学习网络 (CLLNs) 是线性的,这限制了它们对模拟机器学习的能力.
- 为增强功能而将非线性元素集成到CLLN中仍然未被探索.
研究的目的:
- 引入和研究一个非线性对比的本地学习网络 (CLLN).
- 探索将非线性元素纳入电子学习网络的可行性和实用性.
- 证明非线性CLLN在线性系统中难以处理的任务中的学习能力.
主要方法:
- 基于晶体管的自调节非线性电阻元件的模拟电子网络的开发.
- 为非线性CLLN实施一个去中心化的系统架构.
- 测试网络学习非线性任务的能力,包括XOR和非线性回归,无需外部计算机辅助.
主要成果:
- 非线性CLLN成功地学习了与线性系统无法实现的任务,例如XOR和非线性回归.
- 这个去中心化的系统表现出减少错误的模式 (平均值,斜率,曲率),类似于人工神经网络中的光谱偏差.
- 该电路证明了对损坏的强度,快速重新训练 (秒),以及超低能耗 (每个晶体管的picojoules).
结论:
- 非线性CLLN为快速,高效和耐故障的模拟机器学习硬件提供了一条途径.
- 开发的系统显示了低功耗,高性能边缘计算应用的巨大潜力.
- 可扩展的可制造性和研究新兴学习是这项技术的有希望的途径.
相关概念视频
01:26Classification of Systems-I
179
Linearity is a system property characterized by a direct input-output relationship, combining homogeneity and additivity.
Homogeneity dictates that if an input x(t) is multiplied by a constant c, the output y(t) is multiplied by the same constant. Mathematically, this is expressed as:
Homogeneity dictates that if an input x(t) is multiplied by a constant c, the output y(t) is multiplied by the same constant. Mathematically, this is expressed as:
179
01:25Neural Circuits
1.1K
Neural circuits and neuronal pools are two of the main structures found in the nervous system. Neural circuits are networks of neurons that work together to carry out a specific task or process. They consist of interconnected neurons and glial cells, which provide structural and metabolic support.
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
Neuronal pools are collections of nerve cells with similar functions and interact through chemical and electrical signals. These pools include both interneurons (the central neural circuit nodes that...
1.1K
01:26Linear Approximation in Frequency Domain
89
Linear systems are characterized by two main properties: superposition and homogeneity. Superposition allows the response to multiple inputs to be the sum of the responses to each individual input. Homogeneity ensures that scaling an input by a scalar results in the response being scaled by the same scalar.
In contrast, nonlinear systems do not inherently possess these properties. However, for small deviations around an operating point, a nonlinear system can often be approximated as linear....
In contrast, nonlinear systems do not inherently possess these properties. However, for small deviations around an operating point, a nonlinear system can often be approximated as linear....
89
01:26Feedback control systems
304
Feedback control systems are categorized in various ways based on their design, analysis, and signal types.
Linear feedback systems are theoretical models that simplify analysis and design. These systems operate under the principle that their output is directly proportional to their input within certain ranges. For instance, an amplifier in a control system behaves linearly as long as the input signal remains within a specific range. However, most physical systems exhibit inherent nonlinearity...
Linear feedback systems are theoretical models that simplify analysis and design. These systems operate under the principle that their output is directly proportional to their input within certain ranges. For instance, an amplifier in a control system behaves linearly as long as the input signal remains within a specific range. However, most physical systems exhibit inherent nonlinearity...
304
01:30Current Growth And Decay In RL Circuits
3.8K
The current growth and decay in RL circuits can be understood by considering a series RL circuit consisting of a resistor, an inductor, a constant source of emf, and two switches. When the first switch is closed, the circuit is equivalent to a single-loop circuit consisting of a resistor and an inductor connected to a source of emf. In this case, the source of emf produces a current in the circuit. If there were no self-inductance in the circuit, the current would rise immediately to a steady...
3.8K
01:26Signal and System
643
A signal x(t) is a set of data or a time function representing a variable of interest. Signals typically convey information about a phenomenon, such as atmospheric temperature, humidity, human voice, television images, a dog's bark, or birdsongs. More generally, a signal can be a function of more than one independent variable. For instance, images depend on horizontal and vertical positions and can be regarded as two-dimensional signals. However, this text will focus on one-dimensional...
643