Occasional reprint: Summary of AI meetings (by NTU Zhou Zhihua)

Occasional reprint: Summary of AI meetings (by NTU Zhou Zhihua)

Description: Purely personal view, for reference only. Tier-1 's column is more complete, the tier-2 is not too full, tier-3 is not whole.

Alphabetical order of the same points. Not very strictly speaking, tier-1 can be enviable, tier-2 is can make

People respect, because of the AI related meetings very much, so can be included in the tier-3 is also good

Tier-1:

IJCAI (1+): International Joint Conference on Artificial Intelligence aaai (1): National Conference on Artificial Intellig ence

COLT (1): Annual Conference on computational learning theory

CVPR (1): IEEE International Conference on Computer Vision and Pattern recognition

ICCV (1): IEEE International Conference on computer Vision

ICML (1): International Conference on machine learning

NIPS (1): Annual Conference on neural information processing Systems

ACL (N): Annual meeting of the Association for Computational Linguistics

KR (N): International Conference on principles of knowledge representation and reasoning

SIGIR: Annual International ACM Sigir Conference on the and development in information retrieval

SIGKDD (n): ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

UAI (N): International Conference on uncertainty in Artificial Intelligence

*impact factor (according to CiteSeer 03):

ijcai:1.82 (Top 4.09)

aaai:1.49 (Top 9.17%)

colt:1.49 (Top 9.25%)

iccv:1.78 (Top 4.75%)

icml:2.12 (Top 1.88%)

nips:1.06 (Top 20.96%)

acl:1.44 (Top 10.07%)

kr:1.76 (Top 4.99%)

sigir:1.10 (Top 19.08%)

average:1.56 (Top 8.02%)

Ijcai (1+): Ai's best comprehensive conference, starting in 1969, opens every two years, odd-numbered years. Because the AI is too big, so although each session basically can record more than 100 (now more than 200), but in each field is not a few, like Achine Learning, computer vision such a large area of about 10 each time, so the difficulty is very large. But from the hiring rate is not too low, basically about 20%, because the insiders will weigh the weight, do not want to waste reviewer time. Recent articles from mainland China to international conferences are like a tidal wave, and because there are few research groups at home, many conferences in complain say China's low-quality articles are a serious impediment to PC productivity. In this case, it is estimated that the recruitment rate for international conferences will be lowered in the years to come. In addition, the former IJCAI was not poster, beginning in 03, in order to reduce the number of people killed by manslaughter, the increase of 2 pages of paper poster. It is worth mentioning that Ijcai is a "IJCAI Inc" that looks like a company. The host (not actually a company, in fact, a foundation), a few awards at each meeting, the two most important of which are IJCAI Excellence Award and Computer & Thoughts Award, the former is the Lifetime achievement award , each time a person, is basically the AI's highest award (interestingly, the 6 who took the Turing Award in the AI industry, 2 have not received this award), the latter is awarded to the young scientists under the age of 35, each time a person. The award-winning speeches of these two prizes are a major event for each IJCAI. In addition, IJCAI PC member equivalent to other meetings of the area chair, power is very large, because the PC member to find reviewer to audit, and not like the General Conference PC member actually It's reviewer. In order to restrict this power, IJCAI's review process is to allocate 2-bit PC member per article, Primary PC member to find 3-bit reviewer, second PC member to find one.

AAAI (1): Annual Meeting of the American Society of Artificial Intelligence aaai. is a good meeting, but its grade is not stable, can give 1 +, can also give 1-or, in general, I give it "1″." This is because its opening is entirely subject to IJCAI restrictions: open every year, but if this year's IJCAI is held in North America, then stop. So, even years because there is no ijcai, it is the best AI comprehensive conference, but because the appeal is smaller than the IJCAI, especially the Europeans hold aaai field is much less than IJCAI (in fact, Asians are also), so than IJCAI or a little weaker, basically between 1 and 1+; In odd-numbered years, if IJCAI is not in North America, AAAI naturally becomes a lower-level conference than IJCAI (1-or two), for example, in 2005 there were both Ijcai and AAAI, and the coordination of the meetings was conducted, Make IJCAI of the hiring notice time than AAAI deadline early so few days, so IJCAI lost article can be cast to AAAI. IJCAI PC Chair in the review also has been urging, said everyone must be fast, because AAAI There has been worrying about IJCAI's hiring notice late AAAI on the trouble.

COLT (1): This is the best meeting of computational Learning Theory, hosted by ACM, held annually. Computational learning theory can basically be seen as a cross between theoretical computer science and machine learning, so this will be seen by some as theoretical computer science rather than AI. A friend of mine made a wonderful depiction of it in one sentence: "A small group of mathematicians in a meeting." Because the field of colt is relatively small, the annual meeting is basically those people. Incidentally, this is an interesting thing, because the recent meetings in the country too much too many, and many meetings are lncs/lnai out of the proceedings, Lncs/lnai basically has been smelly, but unfortunately, there are some good meetings in Lncs/lnai, such as Colt.

CVPR (1): One of the best meetings of computer vision and Pattern recognition, IEEE hosted, held annually. Although the topic has computer vision, but the individual thinks its pattern recognition taste is more heavier. In fact, it should be the best conference for pattern recognition, and in terms of computer vision, there is a ICCV equivalent. The IEEE has always had a tendency to turn it into a "grand event", which has historically been a "grand event" from the quality. CVPR not good to go this way. There have been a lot of articles in these years. Recently responsible for the CVPR meeting of the TC chair letter said, for this community, let good people are killed than the bad guys are worse, so we want to reduce the chance of the good people being killed? So I guess the cvpr is going to be expanded next year or after.

ICCV (1): Introduced CVPR when said, computer vision is one of the best. IEEE sponsored, held annually.

ICML (1): One of the best sessions in machine learning. It is now hosted by IMLS and held annually. See the introduction to Nips.

NIPS (1): One of the best sessions on Neural computing, hosted by NIPS, is held annually. It is worth noting that the annual venue is the same, formerly the United States of Denver, is now Vancouver, Canada; And it's the end of the year, the 2nd year after the opening of the proceedings, that is to say, NIPS ' 05 of the proceedings are 06 out. The name of the conference "Advances in Neural information processing Systems", so, unlike the "standard" machine learning conferences such as ICML\ECML, Nips has a considerable portion of neuroscience content, and machine learning has a Distance from the set. But because the main content of the conference is machine learning, or close relationship with machine learning, many people regard nips as one of the best machine learning conferences. This meeting basically controls in Michael Jordan's grandchildren hand, therefore to the Jordan Department's person, the hair nips is not difficult, some not very strong work also can send up, but to this circle outside the person, wants to send an article really is very difficult, because leaves "the outsider" the incision is very small. So for people outside of the Jordan system, it's harder to make nips than ICML. In other words, ICML is more open, the influence of the small circle is not as large as nips, so North America and Europe are recognized, and nips some people (especially some Europeans, including some people) are determined not to contribute. This is certainly not a good thing for the Conference itself, but because the Jordan system is so powerful it doesn't seem to care too much. Recently, the IMLS (International Society for Machine Learning) re-elected, including those who have been eligible for nomination for nearly three years in Icml\ecml\colt, Nips was excluded. Anyway, this is a very good meeting.

ACL (n): The best meeting of computational linguistics/Natural Language processing, the ACL (Association of Computational Linguistics) is held annually.

KR: One of the best meetings in knowledge representation and reasoning, in fact one of the best meetings of traditional AI (logic-based AI). KR Inc, now an even-numbered year open.

Sigir: The best conference on Information retrieval, hosted by ACM, opens every year. This will now be a small circle of gas more and more heavy. Information retrieval should not be AI, but because there is more machine learning here, and even a little bit of machine learning Application conference flavor in recent years, so put it in.

SIGKDD: The best conference on data Mining, hosted by ACM, opens every year. The conference is a short one, after all, data mining is just a penis and even a little nephew compared to other fields. A few years ago it was difficult to put it in the tier-1, on the one hand, reputation is far less than the other top conference loud, on the other hand is relatively easy to be hired. But now it is listed in the tier-1 should be a matter of no doubt.

Uai: The name is "uncertainty in artificial intelligence", involving the expression \ Reasoning \ Learning and many other aspects, Auai (Association of UAI), held every year.

Tier-2:

Aamas: International Joint Conference on autonomous Agents and Multiagent Systems

ECCV: European Conference on Computer Vision

ECML: European Conference on machine learning

ICDM: IEEE International Conference on Data Mining

SDM: SIAM International Conference on Data Mining

Icaps (2): International Conference on automated planning and scheduling

ICCBR (2): International Conference on case-based reasoning

Colling (2): International Conference on computational linguistics

Ecai (2): European Conference on Artificial Intelligence

ALT (International): Conference on algorithmic learning theory

EMNLP (+): Conference on empirical Methods in Natural Language processing

ILP (AS): International Conference on inductive Logic programming

PKDD (AS): European Conference on principles and practice of knowledge Discovery in Databases

*impact factor (according to CiteSeer 03):

eccv:1.58 (top 7.20)

ecml:0.83 (Top 30.63)

icdm:0.35 (Top 59.86)

iccbr:0.72 (Top 36.69)

ecai:0.69 (Top 38.49)

alt:0.63 (Top 42.91)

ilp:1.06 (Top 20.80)

pkdd:0.50 (Top 51.26)

average:0.80 (Top 32.02%)

Aamas: The best meeting of the agent. But now the agent is a general concept, almost all AI-related meetings have this content, so the trend of aamas decline is very obvious.

ECCV: Computer vision is second only to the ICCV conference, as this area is developing very quickly and it is possible to upgrade to 1-go.

ECML: Machine learning is second only to ICML, and Europeans are keen to cheer, some think it's already 1. I keep it a little bit, and still put it in the. Because machine learning is developing rapidly, the reputation of this conference is very obvious.

ICDM: Data mining is second only to SIGKDD, and is now equivalent to SDM. This will only be 5 years old, and the pace of ascent is staggering. A few years ago ICDM was not as good as PAKDD, and now it has opened up a great distance.

SDM: Data mining is second only to SIGKDD's conference, and ICDM is now comparable. The foundation of Siam is very thick, but the influence in CS is smaller than that of ACM and IEEE, and SDM is seen to be icdm over, but at least it is still quite.

Icaps (2): The best meeting on AI planning is a merger of previous international and European planning conferences. Because this field is becoming deserted, the impact is smaller than before.

ICCBR (2): The best meeting of case-based reasoning. Because the field is not too big, and has been half cold not hot, so always stay on the 2.

Colling (2): Computational linguistics/Natural Language processing is second only to ACL, but the gap with ACLs is much larger than ICCV-ECCV and ICML-ECML.

Ecai (2): European Artificial Intelligence Conference, a long history, but because of the ijcai/aaai pressure, it is difficult to go up.

ALT (Colt): A bit like the tier-2 version of the book, but because the calculation of learning theory of people do not have much, do a good number to the number of the group, basically to colt go, so there is a lot of Alt is not the content of computational learning theory.

EMNLP: A good meeting in computational linguistics/natural Language processing. Some people think it's quite the same as colling, but I think it's going to be a little weaker.

ILP (+): Inductive logic Program design best meeting. But because there are ILP in many other meetings, it can only keep 2-to-one position.

PKDD (+): Data Mining conference in Europe, currently in the 4th row of Data Mining conference. The Europeans wanted to lift it up, so they had been bundled with ECML all these years, hoping to borrow ECML to bring it up. But because of ICDM and SDM, it's unlikely. So this year's PKDD and ECML although still open together, but has been independently peer review (previously can be cast two at the same time, the author can declare priority by which will consider, if ECML can not also be pkdd accept).

Tier-3:

ACCV: Asian Conference on Computer Vision

DS (International): Conference on Discovery

Ecir: European Conference on IR

Ictai: IEEE International Conference on Tools with Artificial Intelligence

PAKDD: Pacific-asia Conference on Knowledge Discovery and Data Mining

ICANN: International Conference on Artificial Neural Networks

Ajcai (3): Australian Joint Conference on Artificial Intelligence

CAI (3): Canadian Conference on Artificial Intelligence

CEC (3): IEEE Congress on Evolutionary Computation

Fuzz-ieee (3): IEEE International Conference on Fuzzy Systems

GECCO (3): Genetic and Evolutionary computation Conference

ICASSP (3): International Conference on Acoustics, Speech, and Signal processing

ICIP (3): International Conference on Image processing

ICPR (3): International Conference on Pattern recognition

Iea/aie (3): International Conference on Industrial and Engineering applications of Artificial Intelligence and Expert Sys TEMs

IJCNN (3): International Joint Conference on Neural Networks

IJNLP (3): International Joint Conference on Natural Language processing

Pricai (3): Pacific-rim International Conference on Artificial Intelligence

*impact factor (according to CiteSeer 03):

accv:0.42 (Top 55.61%)

ictai:0.25 (Top 69.86)

pakdd:0.30 (Top 65.60)

icann:0.27 (Top 67.73)

ajcai:0.16 (Top 79.44)

cai:0.26 (Top 68.87)

icip:0.50 (Top 50.20)

iea/aie:0.09 (Top 87.79)

pricai:0.19 (Top 76.33)

average:0.27 (Top 68.3%)

ACCV: Asia's computer Vision Conference is good at the Asia-Pacific Conference.

DS: A Japanese-initiated conference that is close to data mining.

ECIR: The European Information Retrieval conference, a few years ago, was just a British information retrieval conference.

Ictai: IEEE's most important artificial intelligence conference, partial application, is a typical of the IEEE run rotten. The former quality is still good, but the longer the reputation is worse, the worse it seems to continue to decline, now in fact, is not too much to stay.

PAKDD: Asia-Pacific Data Mining Conference, currently ranked 5th in the Data Mining conference.

ICANN: European neural Network conferences, from quality, are the best in neural networking meetings, but people in this area do not value meetings, and in this area it is less important than IJCNN.

Ajcai (3): Australia's comprehensive AI conference, at the national/regional level AI conference is good.

CAI (3): Canada's comprehensive AI conference is good at the national/regional AI Conference.

CEC (3): One of the most important conferences in evolutionary computing, the event type. Ijcnn/cec/fuzz-ieee these three meetings are the most important meetings of computational intelligence or soft computing, and they often open together, called Wcci (World Congress on Computational Intelligence). But this field and other branches of CS is not the same, it is similar to other disciplines, only pay attention to journal, do not attach importance to the conference, so the hiring rate is often around 85%, the article has quality very high paper, but also the beginner's work.

Fuzz-ieee (3): The most important meeting of the Blur, the event type, see the introduction of CEC.

GECCO (3): One of the most important conferences in evolutionary computing, comparable to the CEC, the event type.

ICASSP (3): One of the most important speech-related meetings, the people in this area are not very care meetings.

ICIP (3): One of the most famous conferences in image processing, the event type.

ICPR (3): One of the most famous meetings in pattern recognition, the event type.

Iea/aie (3): Artificial intelligence Application Conference. General Conference nomination of excellent papers usually only a few articles, the nomination is already a very high honor, this will be very interesting, each time 1, 20 outstanding paper nominations, specialized in several sessions to do the nomination paper report, it is very lively.

IJCNN (3): The most important meeting and event type of neural network, see the introduction of CEC.

IJNLP (3): A relatively famous conference in computational linguistics/natural Language processing.

Pricai (3): Asia-Pacific Integrated AI conference, although the history is not short, but because it is better or more than a comprehensive conference too much, so it is difficult to rise.

The list is just to help newcomers become familiar with the field, and the ratings or ratings given are personal opinions and are for informational purposes only. In particular, the following are:

1. Tier-1 Conference on the article is not necessarily better than tier-3, can only say that the former average level is higher.

2. The quality of the research is not determined by where it was published, but by the high-grade place just to make it easier for the peer to take notice of the work. The 1 articles cited 10 times at the Tier-3 Conference may be more valuable than the 10 cited 0 articles at the TIER-1 conference. Therefore, the number of top conference articles does not make much sense, it is important to peer evaluation and recognition degree.

3. A lot of classic work is not published in high-grade publishing source, there are a lot of classic work even published in very low-grade publishing source. For many reasons, I will not dwell on it.

4. After all, the conference is a meeting, because the review time is tight, wrong to kill good people and missed bad people are everywhere, not to mention that there are many students who have just begun to do research on behalf of the Boss Peer review.

5. The reputation of the meeting was not static, and the new meeting might not have a reputation at first, but after a few years the pheasant changed to Phoenix, and the old meeting may have had a good reputation, but it was slipping.

6. Only computer science attaches importance to conference papers, and other disciplines do not take the meeting seriously. But in computer science there are also branches that do not attach much importance to meetings.

7. Politics is ubiquitous. Who your boss is, who you are in, and what unit you're in, these simple factors can have a decisive impact. In other words, the difficulty of publishing people in different environments is not the same. With this in mind, you may have special attention to the article high-level published on the source from the obscure author of the low-level unit (for example, if < computer Journal > published an article by the author of the Pingdingshan Railway Electronic Information Technology Institute, I will read it carefully).

8. The evaluation system has a huge impact. No matter where to make a living, scholars need to meet the evaluation system to a certain extent, otherwise even the way of life is not, but also talk about what to do research. At home, since the evaluation system only attaches importance to journal, some scholars who have done very well have never even cast a meeting. In addition, the funds also have a huge restriction. Many good research groups abroad are often important conferences have articles. But domestic is not, low-grade meetings can also be cast only pay registration fees do not meet, high-grade meetings do not go to the report will have a great negative impact, so can only vote very few meetings. This is the most unfavorable place to do CS research in China. One of my speculations: the appreciation of the renminbi will have a small boost to the domestic CS study (of course, the appreciation of the renminbi for the whole of China is greater than the harm or harm than the advantages of it is difficult to say).

Occasional reprint: Summary of AI meetings (by NTU Zhou Zhihua)