OUTLINE

Specific Purpose

To inform my audience about the algorithm and application of Neural Networks.

Central Idea

A Neural Network is an unsupervised feature learning algorithm (which automatically learn a good representation of the input from unlabeled data), and maybe the best ever known algorithm processing complex data in industry .

Introduction

1. The Neural Networks is a new technology that has been used in many kinds of industry.
2. Ordinary people seldom understand what this technology really do.
3. The speech will answer 3 questions: what is a Neural Network, how it works, and what kind of problem it solves.
4. I will skip most mathematical problems, but introduce the Neural Networks algorithm with an example.

Body

1. A Neural Network is a machine learning algorithm.

   • (Definition of learning by Tom Mitchell) A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E.
   • The key problem in machine learning is to learn the target function, which accepts the input and produces output, to learn the experience E.
   • A Neural Network provide a practical method for learning the target function.

2. A prototypical representation of Neural Networks is Multi-layer Perceptron.

   • A multi-layer network consists of an input layer, a few hidden layers, and an output layer.
   • Each layer consists of some differentiable threshold unit, where the target function works.
   • The BACK-PROPAGATION algorithm learns a multilayer network.

3. Neural Networks are widely used to process complex data.

   • Neural Networks provide a way which allow computers to “learn” from data in case of humans only know the fact but cannot tell why.
   • In industry, Neural Networks have been used to analyze complex data.
     
     • Neural Networks have been used in AlphaGo’s algorithm.
     • Neural Networks have been used to build intelligent robots.

Conclusion

1. A Neural Network is a machine learning algorithm which allow computers to “learn” from data in case of humans only know the fact but cannot tell why.
2. Neural Networks work with Multi-layer Perception to find the nonlinear relationships among data.
3. Neural Networks are widely used to process complex data.

下面是PPT，共10页