Overview
This repository contains materials for a short MATLAB and computational neuroscience course, taught at the University of Pennsylvania during the summers of 2009, 2010, and 2011.
You'll probably want to start by downloding this PDF, which contains the syllabus, course notes, and problem sets.
Lecture slides, code, and other materials mat be found by exploring the rest of the repository.
Using this material
Want to use this material in your course? Feel free! You are welcome to use, modify, redistribute, etc. all of the materials for this course. However, (while we have done our best to produce high quality, accurate, materials) we can make no guarantees that the information or code provided is correct. If you notice an error or have a question, post an issue about it!
Contributing
To add new material to the course, or to modify the existing content, please follow these steps:
- Fork this repository
- Make your modifications on your fork
- Submit a pull request with an informative message describing what you changed
Or, to add an idea or propose a change for someone else to make, create a new issue describing what you're thinking.
-
补充资料：神经计算机 神经计算机 neural computer 的偏差来调节连接权值，称为占学习律: △w。二，占矶;a=勺一vj 在已有的各种学习算法中，特别值得一提的是反传算法，或称BP算法。它是为解决多层ANN中隐单元层的训练问题而提出的。BP算法的基本设计思想是:先从输人开始，经过各中间隐单元层逐层计算其输出，直到算出最终输出，并与目标输出比较，得出偏差。然后把这一偏差反传给输出单元的前
-
Abstract 记忆是通过在多个时间尺度上运行的复杂的耦合进程来存储和保留的。为了理解这些错综复杂的交互网络背后的计算原理，我们构建了一类广泛的突触模型，有效利用生物复杂性，通过保护大量记忆免受重写的不利影响来保存它们。记忆容量几乎与突触的数量成线性关系，这比以前的模型的平方根比例有了很大的改善。这是通过组合多个动态过程来实现的，这些动态过程最初将内存存储在快速变量中，然后逐步将它们转移到较慢的
-
反向传播通过使用计算图在Tensorflow，Torch，Theano等深度学习框架中实现。 更重要的是，理解计算图上的反向传播结合了几种不同的算法及其变体，例如通过时间的backprop和具有共享权重的backprop。 一旦将所有内容转换为计算图，它们仍然是相同的算法 - 只是在计算图上反向传播。 什么是计算图 计算图被定义为有向图，其中节点对应于数学运算。 计算图是表达和评估数学表达式的一种
-
Curated list of awesome neuroscience libraries, software and any content related to the domain. Neuroscience is the study of how the nervous system develops, its structure, and what it does. Neuroscie