laitimes

Talk Preview | Beihang Assistant Professor Zhou No. Yi: Three Yuan Attention: The original Transformer also shopped around?

author:Jiangmen Ventures
Talk Preview | Beihang Assistant Professor Zhou No. Yi: Three Yuan Attention: The original Transformer also shopped around?

This issue is The 351st issue of the TechBeat Artificial Intelligence Community

Online Talk is also the 8th issue of the AI Young Scientists Alliance Wutonghui "Class A Plan" series of sharing.

At 8:00 p.m. Beijing time on October 21 (Thursday), Zhou Haoyi's Talk, an assistant professor at Beijing University of Aeronautics and Astronautics, will start broadcasting in the TechBeat artificial intelligence community on time!

The theme he shared with everyone was: "Triple Attention: Turns out Transformers are also shopping around?" At that time, the research progress and challenges of Transformer's Self-attention modeling introduced ternary comparison will be introduced.

Talk Information

topic:

Ternary Attention:

Turns out transformers are also shopping around?

Guest: Assistant Professor zhou yi of Beihang University

Time: 20:00 Beijing time on October 21 (Thursday).

Where: TechBeat Artificial Intelligence Community

http://www.techbeat.net/

Long press to identify the QR code, one click to complete the appointment!

Talk Preview | Beihang Assistant Professor Zhou No. Yi: Three Yuan Attention: The original Transformer also shopped around?

Talk Outline

Ternary relationships are a very important type of connection in reality, but in the current Transformer's Self-attention modeling, only the correlation between two and two is used for similarity calculations. We wonder if we can introduce ternary comparisons as a relational complement or a regular approach. Let the Transformer model also be able to shop around.

The main sharing is:

(1) Background introduction

(2) Ternary attention mechanism

(3) Sharing of experimental results

(4) Summary and outlook

Talk References

This is the information that will be mentioned in this sharing, it is recommended to preview it in advance!

Project Code:

https://github.com/zhouhaoyi/TripletAttention

Talk and ask questions

Ask questions in the following two ways to get WeChat cash red envelope rewards!

Ways (1)

Participate in the interaction in the [Communication Area] under the Talk interface! Leave your call🤟 and questions 🙋, discuss with more friends, and be answered by the speaker directly flop!

Talk Preview | Beihang Assistant Professor Zhou No. Yi: Three Yuan Attention: The original Transformer also shopped around?

Every time you contribute, we will give you the corresponding i-bean points, and there will be surprise rewards!

Talk Preview | Beihang Assistant Professor Zhou No. Yi: Three Yuan Attention: The original Transformer also shopped around?

Ways (2)

Ask your question directly in the message area of this article, or scan the QR code below to ask a question!

Come and ask questions!

Talk guest introduction

Talk Preview | Beihang Assistant Professor Zhou No. Yi: Three Yuan Attention: The original Transformer also shopped around?

Weekly number benefit

Assistant Professor of Beihang Software College

Beihang Big Data Science and Brain-Computer Intelligence High-tech Innovation Center

Dr. Zhou has been committed to the theory and application of artificial intelligence research, solving real problems in industry, medical and other industries. Published more than 10 high-level papers, won the "AAAI 2021 Best Paper Award", won the first prize of the Ministry of Industry and Information Technology's Industrial Internet Innovation Achievement Transformation Award, the first prize of the "Internet +" College Students Innovation and Entrepreneurship Competition in Beijing Division, the open source project Informer ranked first in the PWC community trend and visited more than 200,000 times, was selected into the World Artificial Intelligence Conference WAIC 2021 "Cloud Sail Award • Rising Star", Beijing "Haiying Star", Beijing Outstanding Graduates, Beihang "Top Ten Graduate Students", A rising star in the field of wisdom sources.

In cooperation with the State Grid of China, the equipment fault prediction algorithm developed by him has been applied in Shandong and other countries, covering more than 1,300 ultra-high voltage core transformers of 110kV to 500kV, assisting in the study and judgment of the working conditions of heavy equipment under load, ensuring the safety of major fixed assets; and continuously cooperating with the Third Hospital of Peking University and Peking University Cancer Hospital to establish early warning models such as COPD exacerbation based on spatio-temporal sequence methods, and the joint project was awarded the Beijing "Capital Translational Medicine Innovation Competition" Excellence Award in 2019.

In 2020, as the person in charge, he participated in the new crown epidemic big data prevention and control analysis task of the high-tech innovation center, the planning and pilot verification of the intelligent multi-point trigger monitoring and early warning platform for infectious diseases of the National Health and Health Commission, etc., to support the national epidemic prevention and control decision-making, and received special letters from the General Office of the State Council and the Beijing Municipal Health Commission.

Personal Homepage:

https://www.linkedin.com/in/haoyi-zhou-54a7a69a/

About "Class A Plan"

The "Class A Program" is an elite program initiated by the Alliance of Ai Young Scientists and Indus Hui to tap and cultivate AI academic research talents and entrepreneurial talents. The "Class A Plan" focuses on doctoral students in the direction of artificial intelligence in the world's top universities, and realizes the in-depth integration of industry, academia, research and politics on the basis of the existing resources of the alliance, and becomes an "accelerator" for future academic and industrial leaders.

At the beginning of the selection of the Class A program, the goal is to be a worldwide outstanding doctoral student and start-up founder, and the "hard conditions" include the age of 20-30 years old, the recognition of the world's top academic conferences, and the scientific and technological content of the entrepreneurial direction. Since the establishment of the "A Class Plan" in 2019 at the World Artificial Intelligence Conference, through three sessions of precipitation, "Class A" has welcomed nearly 50 young scientific and technological talents, all of whom are PhD from the world's top universities, with an average age of 26 years old.

As the deputy secretariat unit of the AI Young Scientists Alliance Wutonghui, in the future, we will continue to invite more outstanding members of the A-class program to share their new work, so stay tuned!

The first issue of sharing review: Recurrent.ai Co-founder Yang Zhilin: Natural Language Understanding Model - XLNet

The second issue of the sharing review: Assistant Professor of Zhejiang University: Causal Reasoning and Stable Learning

The third issue of the sharing review: MIT PhD student Hu Yuanming: Taichi programming language - high-performance sparse visual computing and micro programming

The fourth issue of sharing review: Stanford University postdoctoral fellow Wang Hongwei: a personalized recommendation system assisted by knowledge graphs

The fifth issue of sharing review: Bintong Intelligent CEO Gong Chaohui & CTO Ren Zhongqiang: Planning Technology in Flexible Manufacturing Scenarios

Issue 6 Sharing Review: CMU PhD Student Xin Shu tiara: Non-line-of-sight (NLOS) Imaging

Issue 7 Sharing Review: CMU Dr. Liang Junwei: Pedestrian Trajectory Prediction in the Video

Talk Preview | Beihang Assistant Professor Zhou No. Yi: Three Yuan Attention: The original Transformer also shopped around?

If you also want to be a speaker

Self-nomination / referral

Single-player Talk | Team session | Recording or live | Communicate behind closed doors

Choose from a variety of options!

There are also rewards for the success of recommended speakers

About the TechBeat AI Community

TechBeat (www.techbeat.net), affiliated with Shomen Venture Capital, is a growing community that brings together Chinese AI elites around the world.

We hope to create more professional services and experiences for AI talents, accelerate and accompany their learning and growth.

I look forward to becoming a highland for you to learn the frontier knowledge of AI, share the fertile ground of your latest work, and upgrade and fight monsters on the road to AI advancement!

More details>> TechBeat, a learning and growth community that gathers Global Chinese AI elites

Read on