Do we need competition in academia?

In old-fashioned academia publishing there is a competitive factor omnipresent. The average science-conference has only 50 free-slots, while the number of submissions is 500. The consequence is a rejection-rate of 90%. The same is true for academic journals which are published only 4 times a year and on the other hand there are so much new authors who wants to get published. This mismatch is normally solved with a brutal peer-review process which postulates that some papers are not good enough. The concept of selecting only the top 5% of authors / papers is not totally new, also schools and universities are working with the same principle. So called SAT test are done in US-universities for example.

Why the rigorous quality control is done? Normally this question isn’t be asked, it is accepted as the rule of the game, that only the best of the best are welcomed. The real reason can be explained as a cost-problem. A university like harvard rejects most of the applications because the number of rooms on the harvard-campus is limited. The same is true for conferences and journals. They have limited slot numbers.

The answer to the problem is not to peer-review the papers, the answer is in lowering the costs of a journal. If a paper-based journal which have to be send via snail-mail to the libraries is replaced by a digital version, there is no longer a need for limiting the slots. It is possible to increase the acceptance rate to 100%, that means that every submission is printed in the journal.

The consequence is increasing the published paper. To explain the potential loss in quality, a sidestep to the Wikipedia system is necessary. Normally, Wikipedia works with displacement in mind. A better article about “pathplanning”-topic, replaces a worse one. The wikipedia authors defending their own version in so called edit wars. In academia the situation is different. There is no taxonomy which is limited to the article space, instead the informationflow is endless. Instead of replacing former written paper the edit-wars are grouped around attention. It is the same principle like at twitter and the rule is to increase the number of followers.

The number of potential followers is not limited by absolute numbers. Instead the limit is the number of people on the planet earth. If more people are online, more potential followers are possible. And that is the real game of how academia works. It is not a fight for a certain topic or theory, it is a war around the readership. It is true, that academic writers are involved in a competitive game, but the scarce resource is not a slot in a conference or the acceptance letter of journal, instead academia is in competition with other forms of content like weblogs, movies, youtube-clips and webforums. If a blog-post has a better audience than a pdf paper on arxiv, than academia has lost the war.

Advertisements

What is the reputation of CVPR conference?

https://academia.stackexchange.com/questions/75004/is-it-appropriate-to-show-that-a-conference-paper-is-prestigious-by-comparing-th

In a recently posted question the CVPR conference (Conference on Computer Vision and Pattern Recognition) is called a low quality conference:

“the conference is CVPR, … especially to a non-expert audience”

In another thread of the same forum are also concerns expressed about weak papers:

“I find so much in common in those papers. … For example, half the papers seem like: X Algorithm combined with Y algorithm tested on Z database ” https://academia.stackexchange.com/questions/5073/how-do-we-know-if-something-relevant-is-already-published

But how bad is the reputation really? For answering the question let’s take a look into Google Scholar and type in CVPR for getting the proceedings. The result page consists of 148k entries which shows us, that most of the proceedings are online. That speaks for the conference, and not against. In the next step we try to measure the quality of the papers. The best way to do that is to search for something, which looks interesting. For me, I’m currently interested in the following:

CVPR motion vocabulary

I found two interesting papers:

  1. http://xm2vtsdb.ee.surrey.ac.uk/CVSSP/Publications/papers/Mikolajczyk-CVPR-2008.pdf
  2. https://pdfs.semanticscholar.org/5595/03fdd0237cbc3fc10632e38b2fdd2c5f3b75.pdf

… which both looks promising. So as a result I can say, that the quality evaluation of a conference depends on the query which is asked to the proceedings. Perhaps with some tricks it would be possible to find a dedicated low-quality paper which is perhaps not very innovative, but that is not the intention of a reviewer.

Elsevier is under pressure

http://www.sciencemag.org/news/2017/10/publishers-take-researchgate-court-alleging-massive-copyright-infringement

According to the latest news, the company Researchgate involved Elsevier into a lawsuit. The trigger was a copyright violation of Researchgate, which increases the pressure to Elsevier. So the RELX Group which is located in UK had no other option than to defend themself against the aggression. Researchgate has reached a monopol in the area of distributing academic paper over the internet and is listed according to Alexa.com on place #258 worldwide which is more than sciencedirect and springer.com combined. The danger is high, that Elsevier goes bankrupt if they not act.

The motivation behind the biskly behavior of Researchgate is simple: they want to exploit their monopole and crowd out smaller companies from the market. Additional to Elsevier also the American Chemical Society (ACS) is suffering.

Fedora as the new standard-operating system

Traditional, the Linux community has the problem that they is divided into many different distributions. The user base and the programming power is not concentrated on one system but is spread out in many directions. That is major disadvantage for establishing Linux on the desktop. The solution is simple: all hobby-nerds must change to Fedora Linux and the Problem is solved. Fedora is the free version of the Red Hat operating system which is used by major companies. So it is possible to advocate private and commercial usage at the same time. Fedora for Gamers, amateur programmers, creative artists and end-customers, while Red Hat Enterprise Linux for big companies, small companies and universities.

Like all Linux distributions, Fedora has also many problems. For example the Lyx software which i have installed from the repository isn’t stable. Other software like kdenlive is not available under Fedora. But, as a base-system Fedora is good enough. There is a nice Webbrowser out of the box, which gets security patches regularly and many additonal programs like Office tools, games and C++ compilers are also available. So in my opinion Fedora is the best system for increasing the market share of Linux on the desktop. According to the last survey https://www.netmarketshare.com currently Linux have only a portion of 2,3%, while MS-Windows have 89%. Technical both systems are equal, perhaps Windows is using Linux code in the backend too. But the difference is, that a Linux system can be used for advanced things like as a webserver or as programming workstation.

Recent progress in robotics

https://scholar.google.de/scholar?start=0&q=lstm+%22natural+language+instructions%22&hl=de&as_sdt=0,5

A search query at google scholar for the well known “LSTM neural network” shows, that in the last 1-2 years, the number of papers is exploded. More then 10k papers were published, and perhaps it is more because some of them are behind the paywalls. But LSTM isn’t the only hot-topic in AI, another subject which is also interesting is “language grounding”. Both topics combined together realizing nothing less than the first-working Artificial Intelligence. This kind of software is capable of controlling a robot.

But why is the community so fascinated of LSTM and language grounding? At first, the LSTM-network is the most developed neural network to date. It is more powerful than any other neural network. LSTMs are not so perfect, like manual programmed code and some problems like finding prime numbers is difficult to formulate with LSTMs, but for many daily life problems like speech-recognition, image recognition and event-parsing LSTM is good enough. LSTM is not the same as a neural turing machine, so it is not a wonder-power for solving all computerscience-problems, but it is possible to do remarkable things with it.

The main reason why LSTM networks are often used together with language grounding is, that with natural language it is possible to divide complex problems into smaller ones. I want to give an example: If in a research project the robot should be trained with a neural network, to grasp the ball, move around the obstacle and put it into the basket, perhaps the project will fail. Because it takes to much training steps and the problem space is too big, for finding the right LSTM parameters. But with language grounding it is possible to solve each task separately. The network must only learn how to grasp, how to move and how to ungrasp and then the modules can be combined. Sometimes this concept is called multi-modal learning.

Another side-effect is, that the system, even if it was trained with neural networks, remains in control of the human operator. Without any natural commands, the network is doing nothing. Only if the operator types in “grasp”, the neural network is activated. So the system is not really autonomous which must be stopped by the red emergency button, instead it can communicate with the operator via the language which the LSTM network has learned. That makes it easier for finding problems. And if one subtask is too complex for mastering it with an LSTM network, that part can be programmed manual with normal C++ programming language.

https://arxiv.org/pdf/1710.06280.pdf

“Beating Atari with Natural Language Guided Reinforcement Learning, 2017”, https://arxiv.org/pdf/1704.05539.pdf

In the last paper (Beating Atari) is a project described, which is capable for solving the game “MONTEZUMA’S REVENGE” which was in former DeepLearning projects not solvable by AI. What the researcher has done is combining an Neural Network with language grounding and voila, they get a well trainable and high-intelligent Bot.

Bug: Lyx makes trouble again

http://www.lyx.org/trac/wiki/BugTrackerHome

After the last month, the Lyx software for authoring ebooks had shown serious problems. After editing a bit in a text, the screen freezes. This problem occurred only sometime, but the View mechanism for preview a pdf-paper is completely broken. Round about every 20th execution the screen freezes after pressing the button. And it depends not only on the pdflatex engine, but also the xelatex engine makes trouble. So my conclusion is, that inside the Lyx program at all is something really wrong.

In most cases the freeze is recognized after 2 minutes from the operating system and the system is killed hard via the dialog-box of Fedora. If the document was saved correctly depends on random. In 90% of the cases it was saved before, but i had also some cases in which after open the paper again, my text was lost. So if somebody has a time-critical workflow with important documents like a phd-thesis i can no longer recommend the software. Instead the better option would be MS-Word which is very stable and runs under all operating systems.

AI for tactical manoeuvres?

https://ai.stackexchange.com/questions/4421/how-can-an-ai-make-tactical-manoeuvres

I read recently the question and I want to answer not on stackexchange directly but here in my blog. At first, your previous done research into the direction of GOAP and Behaviour trees is perfect for implementing a game AI. There is nothing which are better suited. To the concrete question of how to implement a “Flanking manoeuvre” with an AI the answer is simple: Versioncontrol. That is the development technique which was done in the famous F.E.A.R. game for programming the AI by hand, and this is also the best-practice method for programming your own AI. Version control means, that the project starts with the command “git init” for building a new “.git” folder from scratch and then the programmers have the job to implement the “edit-compile-run” cycle.

In most cases the workflow is similar to painting of an image. It starts with a sketch. On that sketch 3-4 different manoeuvres are visualized and the programmer have the task to implement this with procedural animation in runnable sourcecode. It takes many manhours. If the programming team is really good, they are using natural language as an intermediate authoring tool, so they can write “Team A is attacking place b from left with a circle of 45 degree. Team B is attacking place b at the same time from right with 90 degree flanking.” Than a parser would generate from the natural language a motion pattern for controlling the units in the game.