Cybermath column NZ Math Society Newsletter April 2017

This column takes a break from its recent heavy focus on publication reform to list a few interesting links more related to mathematical research and other professional issues. It is a partially fenced stream of consciousness, but may be useful all the same.

Laci Babai made a bold claim, which generated substantial publicity, that determining whether two graphs are isomorphic can be solved in quasipolynomial time. Harald Helfgott found a flaw while reading the paper deeply in order to present it to S\'{e}minaire Bourbaki, which I had no idea still existed. Babai retracted the claim on 4 January 2017, and reasserted it after fixing the proof on 7 January 2017. How long would this process have taken under the current journal system — would the error have been spotted at all? (sorry, couldn’t resist that). This is an important theoretical breakthrough and shows how well mathematics can work in the internet age.

Speaking of internet mathematics, there is a  journal of that name, devoted really to the mathematics of complex networks (what we used to call graphs before the marketers took over). Not only is the journal interesting and apparently well run, it uses the new platform  Scholastica (as does Tim Gowers’  Discrete Analysis). Another interesting fact is that the journal was formerly published by one of the traditional publishers (Taylor \& Francis), and they gave it up to the editors (not, however, before charging them for the back issues).

Getting back to mathematics on the internet, Polymath is still active, although generating less publicity than a few years ago. They are currently focusing on Rota’s basis conjecture:  if $B_1, B_2, \dots, B_n$ are disjoint bases of an $n$-dimensional vector space $V$  then there exists disjoint bases $C_1, \dots, C_n$ such that each $C_j$ contains one element from each $B_i$.

The arXiv has become very important to mathematicians. At my urging my university will become a financial supporter. I challenge other readers to get their institutions to do the same, rather than freeload as seems to be NZ policy in so many areas in recent years. Although it is cheap to run per paper, the total cost is nontrivial because there are so many papers. It is a challenge to keep up with new postings, so if you trust recommendation algorithms, try  arXivist (“your personal guide to the arXiv”) or Scirate to navigate it. An alternative is to visualise its million-plus papers as a complex network using Paperscape.

The arXiv idea has recently spread to disciplines with very little preprint tradition. The Center for Open Science has developed a preprint platform used by psychology, engineering, sociology and other fields. Maybe journals will change radically soon, after all.

Springer made many old volumes in its Graduate Texts in Mathematics series available for free download in late 2015. The direct links can be found easily by searching although they have apparently revoked the free deal. If they were serious, presumably they would remove the links.

If you want to attend a mathematical meeting in person rather than do everything via the internet, try MathMeetings.Net which aims to be a complete list.

I recently read (much of — far too many letters and namedropping for my taste to finish all of it) The Autobiography of Bertrand Russell. A controversial figure but certainly a mathematician (for part of his life) who followed his conscience wherever it took him. The American Mathematical Society is awarding the Bertrand Russell Prize every 3 years from 2018, for “research or service contributions of mathematicians or related professionals to promoting good in the world and recognizes (sic) the various ways that mathematics furthers human values.” Thomas Hales has apparently funded the prize. It would be good to see nominations (which close 30 June 2017) from this part of the world.

Of course, political activity by mathematicians can cause problems and muddy reputations, as the recently deceased great mathematician Igor Shafarevich found out. Other nonagenarians who have left us recently, in mathematics or related fields, include Kenneth Arrow, Joseph Keller, Howard Raiffa and Thomas Schelling. Going back a year, there is also Christopher Zeeman (whom I am sure visited NZ sometime), Fields Medallist Klaus Roth, while Felix Browder made it to 89. Best wishes to all readers aiming to make 100 while still doing mathematics! The longest-lived mathematician that I am aware of is Leopold Vietoris who not only lived during three centuries, has quite a few concepts named after him.

New algorithms for matching problems

A preprint with Jacky Lo, just submitted to arXiv and on my publications page. The paper was ultimately inspired by a Christmas party game, which shows that the line between work and play is hard to draw.

Abstract: The standard two-sided and one-sided matching problems, and the closely related school choice problem, have been widely studied from an axiomatic viewpoint. A small number of algorithms dominate the literature. For two-sided matching, the Gale-Shapley algorithm; for one-sided matching, (random) Serial Dictatorship and Probabilistic Serial rule; for school choice, Gale-Shapley and the Boston mechanisms. The main reason for the dominance of these algorithms is their good (worst-case) axiomatic behaviour with respect to notions of efficiency and strategyproofness. However if we shift the focus to fairness, social welfare, tradeoffs between incompatible axioms, and average-case analysis, it is far from clear that these algorithms are optimal. We investigate new algorithms several of which have not appeared (to our knowledge) in the literature before. We give a unified presentation in which algorithms for 2-sided matching yield 1-sided matching algorithms in a systematic way. In addition to axiomatic properties, we investigate agent welfare using both theoretical and computational approaches. We find that some of the new algorithms are worthy of consideration for certain applications. In particular, when considering welfare under truthful preferences, some of the new algorithms outperform the classic ones.

Fair Open Access Principles

I have been working for the last 18 months with a group of talented and committed people to accelerate conversion of subscription journals to open access. There are many barriers, and many pitfalls. For example, so-called “predatory” open access journals that take authors’ money and provide no quality control have gained considerable publicity and must be avoided. Large, inefficent and greedy commercial publishers have attempted to “double-dip” by introducing Hybrid OA. Otherwise well-run open access journals still have high publication charges.

Our aim is to avoid these problems by retaining community control of journals and adhering to high ethical standards. Here is a list of our basic principles, based on the original version introduced by LingOA. This list was developed after extensive discussion and some consultation with other OA advocates such as Peter Suber and Marie Farge. We hope it will be useful in delineating what we see as the ideal way to publish journals. This is not to say that all other ways are necessarily “unfair”, of course, although some of them clearly are!

The Fair Open Access Principles

  1. The journal has a transparent ownership structure, and is controlled by and responsive to the scholarly community.
  2. Authors of articles in the journal retain copyright.
  3. All articles are published open access and an explicit open access licence is used.
  4. Submission and publication is not conditional in any way on the payment of a fee from the author or its employing institution, or on membership of an institution or society.
  5. Any fees paid on behalf of the journal to publishers are low, transparent, and in proportion to the work carried out.

Clarification notes:

  1. This could be ownership by an editorial board, or by a democratically controlled scholarly society, for example. Key points are that the controlling organization, not a commercial publisher, must own the journal title, so that a change of service provider can be achieved without changing the title, and so publishing companies simply compete to offer services to the journal. We strongly recommend that the ownership structure allow for democratic input by the community of readers, authors and referees, in addition to editors, and that procedures for making key decisions about the journal’s future be formally (even legally) specified. We strongly recommend that the governing organization be fully nonprofit (for example, IRS 501 (c) (3) in USA). A for-profit company accountable only to shareholders is not compatible with Principle 1.
  2. The  journals and their publishing house can still propose, among their services, to take care of possible legal issues pertaining to copyright on the author’s behalf, under the author’s oversight. We strongly recommend that reviewers also retain copyright of their reviews, and journals retain ownership of all correspondence and mailing lists compiled on the electronic submission system put at their disposal by the publisher.
  3. Any form of subscription paywall is unacceptable, including “hybrid OA”. We  strongly recommend that the industry standard CC-By licence be used. All content of the journal should be easily accessible from the journal website to anyone with a standard internet connection.
  4. The key idea is that journals be “free at the point of use” by authors and readers. Principle 3 deals with readers and Principle 4 with authors. Compulsory APCs (article publication charges) are not compatible with this principle. Journals should ideally be funded by general contributions from universities and research funders, with these contributions not tied to individual articles or groups of authors. Principle 4 is not compatible with “APC Big Deals”, whereby institutions pay for APCs of their employees but do not contribute to a general fund. Also not compatible is the practice of charging APCs by default to the author’s institution, with waivers for authors who do not have institutional funds. The principle does not preclude voluntary APCs, but requests for these must be unobtrusive and no barrier to publication. APCs must be “opt-in”, never “opt-out”.
  5. “Low” depends on the particulars of each journal, but we strongly recommend an absolute maximum of $1000 per article published or $50 per page for the total expense of any journal, and substantially lower fees in all possible cases. We recommend that an itemized price structure be made public in order to ensure transparency and make the proportionality principle apparent.

 

A rant about Lufthansa

Apologies, a rant is required. I just swore at someone on the phone, which is quite a rare occurrence. Read on to see why, if you dare.

Lufthansa is by far the worst airline I have ever had the misfortune to deal with (I haven’t flown on anything really unsafe, of course, but have flown in Iran on a Tupolev). Not only is their in-flight service embarrassingly behind my usual carriers such as Emirates, Singapore, Air New Zealand, their customer service is a joke. My wife has flown them twice in the last year.

Example 1: The first time she bought an in-class upgrade in order to make a further upgrade using Mileage Plus miles. Then was told that the upgrade was not possible. Nor was a refund and in-class downgrade. She flew in the same seat as before, and after 10 months they have offered to refund 20% of the cost, but still haven’t done it more than 3 weeks after asking for bank details. Each time I enquired it took weeks or months to get an email reply.

Example 2: She was denied boarding for LAX-Munich on a trip from Auckland to Europe (because of a newly enforced rule on passport validity that was not picked up at Auckland), and the other 5 legs of the trip were then cancelled. She had to buy a 1-way ticket home. Note that because my wife never made it to her invited conference talk, she doesn’t want to ask the organizers for any money, and they probably think she is being refunded by the airline. The original was bought through Expedia. After hours talking to Expedia we have been told that Lufthansa will within a few days let us know how much they will refund, and then it will take 10 weeks for a refund. I asked Expedia to transfer me to Lufthansa in order to discuss the extra costs incurred that were not on the ticket. The LH staff member told me this number is not for customers, only travel agents. She refused to give me any contact numbers other than the general US 1-800 number. Not wanting to start all over again and failing to convince her with reasoned argument I yelled and hung up.

I don’t want to make too many jokes about Germans, but there is some really stereotypical behaviour going on here. Is customer service, or basic human flexibility, such as difficult concept to grasp?

I strongly urge anyone reading this to avoid Lufthansa if possible. Other tips – never buy your own ticket if it is supposed to be reimbursed by others, get them to buy it directly; make sure your passport has 3 months validity past the end of your stay in Schengen area. And maybe consider cancellation insurance!

What happens to journals that break away?

Although it is still a relatively rare occurrence, several journal boards have broken away from large commercial publishers. A good list is at the Open Access Directory. These journals usually are required to change their name, because the previous publisher will not relinquish it. They are cut off from the enormous support provided by large commercial publishers (after all their subscription prices are so high, the money is surely being put back into developing better infrastructure, rather than, say enriching shareholders, giving inflated honoraria to editors or paying inefficient support staff). Thus one might expect that these journals would struggle.

I looked at the fortunes of the mathematics journals that have taken this route. Below I list the original title name, the approximate date of the breakaway, the new title and publisher, and citation impact measures taken from 2014 data at eigenfactor.org, and compare them to the results for the original journal. Those measures are EF (size-dependent measure of importance) and AI (analogous to Impact Factor, but based on the same kind of reasoning as underlies PageRank – not all citations are equal). Each has a maximum value of 100. These are of course not the only measures one could use. I also list CE, the 2013 cost-effectiveness rating from journalprices.com (essentially, subscription cost per citation) – the smaller the better.

Old: Journal of Logic Programming (Elsevier), changed name more than once to Journal of Logical and Algebraic Methods in Programming, still publishing, EF = 0, AI = 0, CE = 84.73
New: (1999) Theory and Practice of Logic Programming (Cambridge), EF = 31, AI = 40, 42.33

Old: Machine Learning (Springer), EF = 77, AI = 92, CE = 27.01
New: (2001) Journal of Machine Learning Research (diamond OA), EF = 94, AI = 97, CE = 0.0

Old: Topology and Its Applications (Elsevier), still publishing, EF = 78, AI = 33, CE = 32.34
New: (2001) Algebraic and Geometric Topology (Math Sciences Press), EF = 77, AI = 77, CE = 3.67

EDIT: I received email from Alex Scorpan saying: “The facts are that AGT was born by splitting off from “Geometry & Topology”. The resignation of the board of “Topology and its Applications” may have occurred at the same time, may have involved people on the board of AGT, and may have involved the same ethos that moved the founders of GT and AGT, but otherwise the two events are not connected.” Alex has edited the OAD wiki to fix this. I haven’t looked into the question any further.

Old: Journal of Algorithms (Elsevier), stopped publishing after 6-7 years.
New: (2003) Transactions on Algorithms (ACM), EF = 60, AI = 76, CE = 5.57

Old: Topology (Elsevier), stopped publishing after 6 years
New: (2006) Journal of Topology (Oxford), EF = 70, AI = 93, CE = 39.24

Old: K-Theory (Springer), stopped publishing very soon, and archives disappeared.
New: (2007) Annals of K-Theory (Math Sciences Press) (after an intermediate change to Journal of K-Theory (Cambridge), EF = 59, AI = 79, CE = 102.47), too new for EF, AI

Old: Journal of Philosophical Logic (Springer), still publishing, no EF, AI or CE listed (the website lists only “interim editors”)
New: (2007) Review of Symbolic Logic (Cambridge) EF = 35, AI = 49, CE = 222.58

It seems clear that the new journals are doing considerably better than the old ones overall. I wonder whether the idea often touted by radical leftist OA advocates that large commercial publishers don’t add much value could have a grain of truth in it.

A blockage cleared

After a very long time, some (revised versions of) papers have emerged, two on distance rationalizability of voting rules with Benjamin Hadjibeyli, one on signed networks with Samin Aref, and one on “wisdom of crowds” with Patrick Girard and Valery Pavlov.

There is a long list of things still to do, but this is a big relief. These papers had gone through substantial revisions and years of work in some cases. There is also a recent paper on legislation networks with Neda Sakhaee and Golbon Zakeri.

My papers can always be found on my publications page.

Otello

I saw a concert version of this tonight with Simon O’Neill in the title role. While the plot is quite weird and distasteful in places, the orchestra and singing were excellent. I saw Simon O’Neill in 1997 (I think) in a small role in the play Master Class in Auckland. He has certainly gone up in the world!

COMSOC 2016

I attended this biennial meeting for the fourth consecutive time. Attendance in Toulouse was substantially larger than previously. Organization was excellently led by Umberto Grandi.

Toulouse was busy with tourists augmented by fans of Euro 2016 football teams and evenings were very noisy downtown. The city seems like a pleasant, slightly provincial place.

There were many interesting talks but I find it hard to maintain concentration. The poster sessions were surprisingly interesting and perhaps in future there should be much more time spent on these, and shorter talks given.