On May 5, 2018, I watched the online open class of Tech Nuggets with anticipation. The topic of AI this time is AI. I am currently learning Al, and I am learning about image super-resolution and NLP. In the long dark night , Nuggets - a technical community that guides me and lights the way forward. This class made me feel very much. I, who have always been unknown in Zhihu, where the big guys gather, have posted on Zhihu. The first article "Why is the society developing faster and faster? ——Artificial intelligence may really make human beings eventually perish. The website is as follows: https://zhuanlan.zhihu.com/p/36541542 Comments are welcome.
———————————Call me gorgeous dividing line————————
The Nuggets invited two technical leaders, one is Mr. Shao Jieshao from Qiniu Cloud AI Lab, and the other is Mr. Peng Yaopeng from Qiniu Cloud AI Lab.
Mr. Shao mainly talked about the current research field of AI and the development and application of various aspects, such as natural language processing (natural language processing) knowledge representation (knowledge representation) and automated reasoning (automated reasoning).
Machine learning (machine learning), computer vision (computer vision, robotics, robotics), and at the same time give some learning suggestions to our friends who are new to AI, such as: (1) Don't wait until you have mastered all the relevant mathematical knowledge before starting, because mathematics There are a lot of things, and it is difficult, and the energy is limited. When you need to use mathematics, you can make up for it.
(2) Don't collect too much learning materials, because the online materials are complicated, the quality is uneven, the knowledge points are scattered, not systematic, time is precious, and energy is limited.
(3) Do more hands-on and more practical operations. After all, computer is a subject with strong practical operation. Next, Mr. Shao will tell us how to use learning algorithms to generate models from data, such as: Spam filters Search ranking Click through rate predict Recommendations Speech recognition Machine translation Face detection Image classification
At the same time, introducing the algorithm to us requires specific analysis of specific problems, because the function of the algorithm is to play its maximum role in a specific situation, and different algorithms are suitable for different situations and systems. Teacher Shao also introduced the content of image processing, because it is widely used in daily life, such as video surveillance, image printing, medical image processing, satellite imaging, military reconnaissance and so on.
Machine Learning: Using learning algorithms to generate models from data. Simply put, according to the written program (machine algorithm), according to a large amount of data, a model is generated. In fact, the deep learning neural network is commonly used now. The lecturer also talked about an example: for example, you often receive spam, and next time you receive it, you will analyze the received spam to determine whether it is spam.
Machine Learning: Generalization (analyzing new data based on existing data), algorithm preference (different models, problems, applications match different algorithms).
Regarding machine learning, K-nearest neighbors are also used to realize an image recognition. Solid mathematical knowledge is required.
The teacher briefly talked about some of the main contents of machine learning, including loss function, regular term, optimization, hyperparameters, etc. Recommended books are: Nick "A Brief History of Artificial Intelligence" Miroslav Kubat "Introduction to Machine Learning" Zhou Zhihua "Machine Learning" (Watermelon Book) Aurelien Geron "Hands-on Machine Learning with Scikit-learn & Tensorflow" Ian Goodfellow and other "Deep Learning" (Flower Book) The recommended online courses are Machine Learning Crash Course https://developers.google.com/machine-learning/crash-course/ National Taiwan University Professor Li Hongyi http://speech.ee.ntu.edu.tw/ ~tlkagk/courses.html Professor Wu Enda http://mooc.study.163.com/smartSpec/detail/1001319001.htm Stanford University cs231n http://cs231n.stanford.edu/ Stanford University cs224n http://web.stanford .edu/class/cs224n/
The final question 1. Do you recommend starting with deep learning? 2. Phtroch and TensorFlow are two machine learning libraries, which one is better for learning? 3. Hands-on is very important, how should you practice it? 4. What is the appropriate language for AI development? 5. May I ask the teacher, what advice do you have for the traditional software development industry (C language) and Xingxing artificial intelligence industry (machine learning direction)? 6. What does the teacher think of online machine learning courses, such as coursera, etc. Please tell me 7. What are the optimization schemes from low resolution to high resolution in image processing? (This question was actually raised by me.) At that time, the teacher replied that how did the low-resolution images come from, in fact, noise reduction, etc. or ordinary camera equipment would do. .
———————————Call me gorgeous dividing line————————
Peng Yao Peng mainly talked about three aspects 1. How does AI change our life? Several major applications of artificial intelligence 2. The application of AI in Qiniu Cloud? Introducing the Seven Cloud Bull
Qiyunniu's development achievements in the field of video and intelligence
There is also the core innovation system of Qiyunniu's artificial intelligence laboratory
Including (1) Content review: use artificial intelligence machine vision technology, massive video image data to identify pornography, violence, terrorism, and politics, and ensure the health of the Internet, radio, television, new media, and the government on the content of data dissemination. (2) Eyes of the City: Use artificial intelligence machine vision technology to conduct fast and efficient "detection, recognition and behavior analysis" of "people, objects, and scenes" to meet the needs of users in "identity verification", "intelligent security", "big large-scale image and video retrieval” and other aspects of the scene requirements. Based on the core framework of AI machine learning, the detection and recognition speed is fast. With the increase of the sample size and the amount of learning, the accuracy rate will increase rapidly. (3) Media asset intelligence (4) Broadcast control system (5) DORA daily average tens of billions of intelligent multimedia API platform
3. The daily work of Qiniu cloud AI engineer Type of seven cloud cattle engineer Computer vision algorithm engineer Machine learning platform R&D engineer Big data platform R&D engineer Search engine R&D engineer System architecture engineer Business architecture engineer
Well, that's all, I feel that I have received a lot of goods! Many thanks to the Nuggets and Qiyunniu for this open class! ! !