Tencent advertising algorithm competition can have such charm, not only rely on the bonus, but also from the real scene based on the competition questions, massive data, free super computing resources, as well as the great attention to talents. The Tencent advertising algorithm competition sponsored by Tencent advertising has entered its fourth year, and this year’s competition has been further upgraded in many aspects: < / P > < p > in terms of bonus, this year’s prize pool has been expanded to the level of one million. The champion team will get a high prize of 70000 US dollars, and the second runner up team and the second runner up team will get 10000 US dollars and 5 respectively, The rest of the top 10 teams in the finals will also receive $1000 each. Last year, the vice president of the advertising department of Tsinghua University, including Vice President of the advertising department of Tsinghua University, vice president of the Department of advertising, Mr CEO Huang Shifei and other Tencent executives. At the final scene, the Organizing Committee of Tencent advertising algorithm competition also invited Liu Yuhong, deputy general manager of Tencent Data Platform Department, MI Qi, director of Intel global key account, Xu Xiaomin, general manager of AI basic product center of Tencent cloud products department, as well as a number of technical executives. < p > < p > in terms of resource support, Tencent advertising algorithm competition has introduced more partners, including Tencent cloud AI, Tencent big data, Tencent recruitment, Tencent university cooperation and Intel. At the same time, Tencent conference provides remote collaboration, online conference and live broadcast services for the whole competition. Intel and Ti-One, Tencent cloud intelligent titanium machine learning platform, jointly support the AI algorithm platform. Meanwhile, the competition recommends the use of angel training framework optimized for sparse high-dimensional model. < p > < p > in addition, due to its challenging and interesting nature, the contest itself has become a different landscape. < / P > < p > according to the desensitization data of Tencent’s real advertising business, the machine learning technology is used to accurately predict the user’s attributes when the user data is sparse. It can be said that such a contest has a strong practical significance, which can not only ensure the privacy and security of users’ data, but also solve the practical problem of the advertiser’s own data sparsity in the launch. In addition, “dyg” also proposes the method and model of “fusion probability distribution”, which uses multi-layer transformer to fuse probability distribution and semantic representation of Bert, which effectively improves the performance. < / P > < p > < p > the team that won the second place in the “mountain has wood” team also wanted to directly adopt Bert in the pre training model stage. However, considering the factors such as training time and debugging cost, the team finally changed a way – imitating Bert to realize hierarchical tag information injection, and then through target Encoding, multi model fusion and other operations constantly refresh the competition results, and finally won the second place in this competition. < / P > < p > among the third to tenth ranked teams, although some teams indicated that they had considered using Bert, they still adopted commonly used LSTM and other technologies in the end due to the problem of computational complexity. In this regard, Jiang Jie said in a qubit interview that the calculation amount of the first two teams is not large, because they have carried out a strong compression on the whole parameter after embedding, that is to say, they have done a screening. < p > < p > no matter in terms of manpower, material resources, financial resources and energy, Tencent’s advertising has made a lot of investment in this competition every year, and this year’s epidemic situation has made it even more difficult. < / P > < p > interview through train – excellent teams can get free of written examination. Such cases are not rare in the past, but for the new students facing employment difficulties this year, joining Tencent through technical competition is a very attractive olive branch. < p > < p > welcome these players who have won good places and can continue to participate in the algorithm competition next year. We are also looking forward to more players joining Tencent through the algorithm contest. < / P > < p > this year, the number of applicants for Tencent’s advertising algorithm competition has reached 11000, and the algorithm competition of this scale is also the top in the world. < / P > < p > the original intention of Tencent advertising to hold Tencent advertising algorithm competition for four years is to recruit talents from all over the world and create a platform for technical exchange and sharing. < p > < p > of course, Tencent advertising, which attaches so much importance to talents, explores technology not only in Tencent advertising algorithm contest, an annual technology event. < / P > < p > the project is aimed at opening up the technical challenges in Tencent’s advertising business, building a platform for industry university research cooperation, jointly exploring frontier issues affecting the long-term development of advertising technology, and cultivating outstanding talents. < p > < p > since 2018, Tencent’s annual “rhinoceros bird special research program” has also attracted extensive attention and active participation from scholars at home and abroad. More than 20 colleges and universities have started special cooperation with Tencent advertising through layer by layer screening, and have carried out close scientific research cooperation with advertising R & D teams facing real problems of advertising and actual business needs. < p > < p > this year’s “rhinoceros bird special research project” has set up eight research topics focusing on machine learning, recommendation system and natural language processing. To some extent, it also reflects the technical exploration direction of Tencent advertising. < / P > < p > more than one million research fund support. The project will be open to the participation of teachers and students. It is a valuable opportunity to go to Tencent for internship, and carry out research work under the joint guidance of R & D personnel and school tutors. < p > < p > in addition, Tencent advertising cooperated with Tencent universities to build a live broadcast activity of academic and industrial circles, wiztalk, and shared with marketers and technicians the achievements of rhinoceros special cooperation over the years in an easy to understand tone. < / P > < p > from the development of guangdiantong team in 2011, Tencent’s advertising technology undoubtedly belongs to the first echelon. But we certainly will not stop here, and we need to explore the most cutting-edge technical propositions with a searching attitude. And the algorithm contest is a great way. Talent is Tencent’s most valuable wealth. Only by allowing more talents to join Tencent and come to Tencent advertising, can Tencent’s advertising platform ability and technical ability be continuously improved. < / P > < p > in fact, if you pay attention to “algorithm” and “goose factory”, you should be familiar with this increasingly influential competition and Tencent advertising behind it. At present, < / P > < p > advertising is the most mature mode of Internet commercialization. As one of the top Internet companies in China, Tencent’s advertising business is also in the forefront of the industry in terms of technical exploration of commercialization: < / P > < p > there are 10 billion dimensions of data urgently needed by AI, complex and diversified effect transformation links, which can continuously pose new challenges, and have industry application scenarios that can produce effects and feedback continuously. < p > < p > in the mid-year online summit of “Tencent advertising, the great change of China show marketing in 2020”, Tencent advertising put forward its new positioning — commercial service center. < p > < p > rooted in Tencent’s unique C to B “super connection” system, Tencent advertising for the first time systematically proposed four connection models of full link digital marketing, namely brand mental connection, transaction transformation connection, private user connection and experience innovation connection. This model is positioned in Tencent’s business service platform for enterprises, and links with the whole platform resources and ecological partners to help enterprises realize the transformation of full link digital marketing and better realize the connection between users and commerce. Before he became Vice President of Tencent advertising at the end of last year, Jiang Jie, who joined Tencent in 2012, was already the general manager of Tencent’s data platform department and vice president of Tencent’s smart retail industry research. However, he has always been self abased as a “new advertising man” in the interview. < / P > < p > after joining Tencent’s advertising, he has carried out a number of integration work in a bold manner, realizing the multi-end unification and upgrading of the delivery end and API ecology, and thus promoting the technical integration of multiple dimensions such as data capability and strategic ability. < p > < p > from the three aspects of “user understanding”, “user touch and influence” and “user transformation and operation”, the advertising technology business of Tencent, which Jiang Jie is responsible for, will also provide comprehensive support for the full link digital marketing of enterprises. < p > < p > the product and technology vision of Tencent advertising aims to connect users and commerce, drive the growth of users in the whole transaction link, and let technology create greater value for business. However, the middle platform of Tencent’s advertising is not planned from a top-down perspective, nor does it call out a slogan of “middle platform” and then build it gradually. < / P > < p > Zhongtai is precipitated step by step in the process of service business. Tencent advertising runs through the “Zhongtai idea” in every service to advertisers. The advertising link is very long. Tencent advertising needs to achieve technological breakthroughs in every link of the whole chain. The optimization based on the full link will also be a lasting battle for the technical team of Tencent advertising. In this regard, Tencent advertising will not sacrifice users’ privacy. On the contrary, Tencent advertising has always implemented Tencent’s vision and mission of “user-oriented, technology oriented” and put the protection of user’s privacy in the most important position. < / P > < p > the title of this algorithm competition is also set up based on the protection of user privacy. It is hoped that through the in-depth understanding of user data characteristics, the accuracy of advertising business can be improved on the basis of legal compliance, and the user experience can be maximized under the premise of protecting user privacy. < p > < p > in the first quarter of 2020 affected by the epidemic, Tencent’s online advertising revenue increased by 32% year-on-year to RMB 17.713 billion, and social and other advertising revenue increased by 47% to RMB 14.592 billion. < / P > < p > we also believe that a rare technology exchange platform like Tencent advertising algorithm competition can provide more talents for Tencent advertising and Tencent, let users and advertising meet naturally in a beautiful scene, and make the commercialization path of the Internet become more intelligent and intelligent. Continue ReadingAmerican companies begin to give up R & D: who should pay for corporate research?