CN C1 Introduction

Course Info

Contents	Percentage
Assignment and Presence	40
Final	60

summary

Lec	Contents
1	Neuroscience Background
2	Neurons, Synapse and Signal Transmission
3,4	Neural Encoding
5,6	Sensory Systems I & II
7,8	Memory Mechanisms and Models I & II
9	Learning Mechanisms
10,11	Supervised Learning
12	Unsupervised Learning
13	Wiring to form network
14	Visual Perception and High-level Cognition
15	BCI and Robotic Cognition
16	Review

Exam

• 掌握大脑神经信息处理和计算的基本机制
• 掌握神经元的基本计算模型及算法
• 掌握不同输入对神经活动的影响
• 掌握主要神经可塑性模型
• 掌握代表性学习算法
• 理解基本的大脑感知和认知机制

Background about CN

CN, From neural-level activities to system-level behaviors and cognition, aims at to understand our brain and to build brain-like computing framework.

Bell–Magendie law：运动纤维、感觉纤维的作用和神经冲动在神经中的运动是单向的。

Nerve Electrical Signal

Bernstein was able to amplify the strength of the signal and record it at a finer timescale, crated the first true observation of the nervous electrical signal.

The current rapidly decrease and them slowly recover to its normal value.

Hodgkin-Huxley Neuron Mode

Both solved the mystery of action potential (AP), describing the equivalent circuit, and a set of complex equations.

McCulloch-Pitts Computation Model

To think of the brain as a computing device following the rule of logic helps understand the thoughts in terms of neural activity, rather than a bag of proteins and chemicals.

AIso the Start of Artificial Intelligence.

AI and Neuroscience become the pillar to each other.

The bottle neck of STOA

• weak robustness and generalization
• weak Flexibility
• Big data, High computation cost, but single, small tasks

Computer Algorithms and Architectur

A computing substrate (von Neumann architecture) is completely detached from algorithms (Turing Machine framework).

Information flow within a vertebrate neuron. Neurons generally use dendrites (blue) to receive information from their presynaptic partners and the axon (red) to send information to their postsynaptic partners. Thus, information flows from dendrites to cell bodies to the axon (black arrows).

Major Themes about CN

Biological Intelligence

Cognitive navigation and Object recognition are easy for animals but hard for robots to mimicking.

• Cognitive computing is to develop a coherent, unified, universal mechanism inspired by the mind’s capabilities.

Brain-Inspired Artificial Intelligenc

• System level： 以⽬标为导向的模拟⼤脑的系统模型, 包括⼤脑使⽤的算法、结构、功能和表达.

  e.g深度学习、强化学习、注意⼒、记忆、连续学习、隐学习(潜意识学习)、 元学习（learning to learn）、迁移学习（transfer learning）、想象与规划等。

• Algorithmic level：实现上述⽬标的计算和过程；

• Implementation level： 更加精确仿⽣的实现机制

  Neuromorphic computing is a set of technologies that seek to mimic biological neural networks to improve the hardware efficiency and robustness of computing systems

Complex Synapse Organization Structure

Neural Circuits Based Perception

The neuronal representation of entire objects is central to high-level visual processing. Object representation

• involves integration of visual features. Ideally the resulting representation is a generalization of the numerous retinal images generated by the same object and of different members of an object category.
• incorporates information from other sensory modalities
• attaches emotional valence and associates the object with the memory of other objects or events.
• can be stored in working memory and recalled in association with other memories.

Sensory Neurons

• Visual
• Temperature
• Pressure
• Touch
• Hearance

Three levels of Cognitive Neuroscience

• 神经元（Neurons）
• 神经环路（Neuron Circuits）
• 脑的⾼级功能和⾏为（Behavior)