If you are interested, please email msrnycrsvp at microsoft.com and say “I want to come” so we can get a count of attendees for refreshments.
The last several years have seen a phenomonal growth in machine learning, such that this earlier post from 2007 is understated. Machine learning jobs aren’t just growing on trees, they are growing everywhere. The core dynamic is a digitizing world, which makes people who know how to use data effectively a very hot commodity. In the present state, anyone reasonably familiar with some machine learning tools and a master’s level of education can get a good job at many companies while Phd students coming out sometimes have bidding wars and many professors have created startups.
Despite this, hiring in good research positions can be challenging. A good research position is one where you can:
- Spend the majority of your time working on research questions that interest.
- Work with other like-minded people.
- For several years.
I see these as critical—research is hard enough that you cannot expect to succeed without devoting the majority of your time. You cannot hope to succeed without personal interest. Other like-minded people are typically necessary in finding the solutions of the hardest problems. And, typically you must work for several years before seeing significant success. There are exceptions to everything, but these criteria are the working norm of successful research I see.
The set of good research positions is expanding, but at a much slower pace than the many applied scientist types of positions. This makes good sense as the pool of people able to do interesting research grows only slowly, and anyone funding this should think quite hard before making the necessary expensive commitment for success.
But, with the above said, what makes a good candidate for a research position? People have many diverse preferences, so I can only speak for myself with any authority. There are several things I do and don’t look for.
- Something new. Any good candidate should have something worth teaching. For a phd candidate, the subject of your research is deeply dependent on your advisor. It is not necessary that you do something different from your advisor’s research direction, but it is necessary that you own (and can speak authoritatively) about a significant advance.
- Something other than papers. It is quite possible to persist indefinitely in academia while only writing papers, but it does not show a real interest in what you are doing beyond survival. Why are you doing it? What is the purpose? Some people code. Some people solve particular applications. There are other things as well, but these make the difference.
- A difficult long-term goal. A goal suggests interest, but more importantly it makes research accumulate. Some people do research without a goal, solving whatever problems happen to pass by that they can solve. Very smart people can do well in research careers with a random walk amongst research problems. But people with a goal can have their research accumulate in a much stronger fashion than a random walk through research problems. I’m not an extremist here—solving off goal problems is fine and desirable, but having a long-term goal makes a long-term difference.
- A portfolio of coauthors. This shows that you are the sort of person able to and interested in working with other people, as is very often necessary for success. This can be particularly difficult for some phd candidates whose advisors expect them to work exclusively with (or for) them. Summer internships are both a strong tradition and a great opportunity here.
- I rarely trust recommendations, because I find them very difficult to interpret. When the candidate selects the writers, the most interesting bit is who the writers are. Letters default positive, but the degree of default varies from writer to writer. Occasionally, a recommendation says something surprising, but do you trust the recommender’s judgement? In some cases yes, but in many cases you do not know the writer.
Meeting the above criteria within the context of a phd is extraordinarily difficult. The good news is that you can “fail” with a job that is better in just about every way
Anytime criteria are discussed, it’s worth asking: should you optimize for them? In another context, Lines of code is a terrible metric to optimize when judging programmer productivity. Here, I believe optimizing for (1), (2), (3), and (4) are all beneficial and worthwhile for phd students.
This year’s ICML had several papers which I want to read through more carefully and understand better.
- Chun-Liang Li, Hsuan-Tien Lin, Condensed Filter Tree for Cost-Sensitive Multi-Label Classification. Several tricks accumulate to give a new approach for addressing cost sensitive multilabel classification.
- Nikos Karampatziakis and Paul Mineiro, Discriminative Features via Generalized Eigenvectors. An efficient, effective eigenvalue solution for supervised learning yields compelling nonlinear performance on several datasets.
- Nir Ailon, Zohar Karnin, Thorsten Joachims, Reducing Dueling Bandits to Cardinal Bandits. An effective method for reducing dueling bandits to normal bandits that extends to contextual situations.
- Pedro Pinheiro, Ronan Collobert, Recurrent Convolutional Neural Networks for Scene Labeling. Image parsing remains a challenge, and this is plausibly a step forward.
- Cicero Dos Santos, Bianca Zadrozny, Learning Character-level Representations for Part-of-Speech Tagging. Word morphology is clearly useful information, and yet almost all ML-for-NLP applications ignore it or hard-code it (by stemming).
- Alekh Agarwal, Daniel Hsu, Satyen Kale, John Langford, Lihong Li, Robert Schapire, Taming the Monster: A Fast and Simple Algorithm for Contextual Bandits. Statistically efficient interactive learning is now computationally feasible. I wish this one had been done in time for the NIPS tutorial
- David Silver, Guy Lever, Nicolas Heess, Thomas Degris, Daan Wierstra, Martin Riedmiller, Deterministic Policy Gradient Algorithms. A reduction in variance from working out the deterministic limit of policy gradient make policy gradient approaches look much more attractive.
Edit: added one that I forgot.
The key reason for this is information: I expect everyone participating in ICML has some baseline interest in how ICML is doing. Everyone involved has personal anecdotal information, but we all understand that a few examples can be highly misleading.
Aside from satisfying everyone’s joint curiousity, I believe this could improve ICML itself. Consider for example reviewing. Every program chair comes in with ideas for how to make reviewing better. Some succeed, but nearly all are forgotten by the next round of program chairs. Making survey information available will help quantify success and correlate it with design decisions.
The key question to ask for this is “who?” The reason why surveys don’t happen more often is that it has been the responsibility of program chairs who are typically badly overloaded. I believe we should address this by shifting the responsibility to a multiyear position, similar to or the same as a webmaster. This may imply a small cost to the community (<$1/participant) for someone’s time to do and record the survey, but I believe it’s a worthwhile cost.
I plan to bring this up with IMLS board in Beijing, but would like to invite any comments or thoughts.
- Rayid Ghani (Chief Scientist at Obama 2012)
- Brian Kingsbury (Speech Recognition @ IBM)
- Jorge Nocedal (who did LBFGS)
We’ve been somewhat disorganized in advertising this. As a consequence, anyone who has not submitted an abstract but would like to do so may send one directly to me (firstname.lastname@example.org title NYASMLS) by Friday March 14. I will forward them to the rest of the committee for consideration.
In recent years, there’s been an explosion of free educational resources that make high-level knowledge and skills accessible to an ever-wider group of people. In your own field, you probably have a good idea of where to look for the answer to any particular question. But outside your areas of expertise, sifting through textbooks, Wikipedia articles, research papers, and online lectures can be bewildering (unless you’re fortunate enough to have a knowledgeable colleague to consult). What are the key concepts in the field, how do they relate to each other, which ones should you learn, and where should you learn them?
Courses are a major vehicle for packaging educational materials for a broad audience. The trouble is that they’re typically meant to be consumed linearly, regardless of your specific background or goals. Also, unless thousands of other people have had the same background and learning goals, there may not even be a course that fits your needs. Recently, we (Roger Grosse and Colorado Reed) have been working on Metacademy, an open-source project to make the structure of a field more explicit and help students formulate personal learning plans.
Metacademy is built around an interconnected web of concepts, each one annotated with a short description, a set of learning goals, a (very rough) time estimate, and pointers to learning resources. The concepts are arranged in a prerequisite graph, which is used to generate a learning plan for a concept. In this way, Metacademy serves as a sort of “package manager for knowledge.”
Metacademy also has wiki-like documents called roadmaps, which briefly overview key concepts in a field and explain why you might want to learn about them; here’s one we wrote for Bayesian machine learning.
Many ingredients of Metacademy are drawn from pre-existing systems, including Khan Academy, saylor.org, Connexions, and many intelligent tutoring systems. We’re not trying to be the first to do any particular thing; rather, we’re trying to build a tool that we personally wanted to exist, and we hope others will find it useful as well.
Granted, if you’re reading this blog, you probably have a decent grasp of most of the concepts we’ve annotated. So how can Metacademy help you? If you’re teaching an applied course and don’t want to re-explain Gibbs sampling, you can simply point your students to the concept on Metacademy. Or, if you’re writing a textbook or teaching a MOOC, Metacademy can help potential students find their way there. Don’t worry about self-promotion: if you’ve written something you think people will find useful, feel free to add a pointer!
We are hoping to expand the content beyond machine learning, and we welcome contributions. You can create a roadmap to help people find their way around a field. We are currently working on a GUI for editing the concepts and the graph connecting them (our current system is based on Github pull requests), and we’ll send an email to our registered users once this system is online. If you find Metacademy useful or want to contribute, let us know at feedback _at_ metacademy _dot_ org.
At NIPS I’m giving a tutorial on Learning to Interact. In essence this is about dealing with causality in a contextual bandit framework. Relative to previous tutorials, I’ll be covering several new results that changed my understanding of the nature of the problem. Note that Judea Pearl and Elias Bareinboim have a tutorial on causality. This might appear similar, but is quite different in practice. Pearl and Bareinboim’s tutorial will be about the general concepts while mine will be about total mastery of the simplest nontrivial case, including code. Luckily, they have the right order. I recommend going to both
I also just released version 7.4 of Vowpal Wabbit. When I was a frustrated learning theorist, I did not understand why people were not using learning reductions to solve problems. I’ve been slowly discovering why with VW, and addressing the issues. One of the issues is that machine learning itself was not automatic enough, while another is that creating a very low overhead process for doing learning reductions is vitally important. These have been addressed well enough that we are starting to see compelling results. Various changes:
- The internal learning reduction interface has been substantially improved. It’s now pretty easy to write new learning reduction. binary.cc provides a good example. This is a very simple reduction which just binarizes the prediction. More improvements are coming, but this is good enough that other people have started contributing reductions.
- Zhen Qin had a very productive internship with Vaclav Petricek at eharmony resulting in several systemic modifications and some new reductions, including:
- A direct hash inversion implementation for use in debugging.
- A holdout system which takes over for progressive validation when multiple passes over data are used. This keeps the printouts ‘honest’.
- An online bootstrap mechanism system which efficiently provides some understanding of prediction variations and which can sometimes effectively trade computational time for increased accuracy via ensembling. This will be discussed at the biglearn workshop at NIPS.
- A top-k reduction which chooses the top-k of any set of base instances.
- Hal Daume has a new implementation of Searn (and Dagger, the codes are unified) which makes structured prediction solutions far more natural. He has optimized this quite thoroughly (exercising the reduction stack in the process), resulting in this pretty graph.
Here, CRF++ is commonly used conditional random field code, SVMstruct is an SVM-style approach to classification, and CRF SGD is an online learning CRF approach. All of these methods use the same features. Fully optimized code is typically rough, but this one is less than 100 lines.
I’m trying to put together a tutorial on these things at NIPS during the workshop break on the 9th and will add details as that resolves for those interested enough to skip out on skiing
Edit: The VW tutorial will take place during the break at the big learning workshop from 1:30pm – 3pm at Harveys Emerald Bay B.
Various news stories have coverage, but the synopsis is that he had a heart attack on Sunday and is survived by his wife Anat and daughter Aviv. There is discussion of creating a memorial fund for them, which I hope comes to fruition, and plan to contribute to.
I will remember Ben as someone who thought carefully and comprehensively about new ways to do things, then fought hard and successfully for what he believed in. It is an ideal we strive for, that Ben accomplished.
Several strong graduates are on the job market this year.
- Alekh Agarwal made the most scalable public learning algorithm as an intern two years ago. He has a deep and broad understanding of optimization and learning as well as the ability and will to make things happen programming-wise. I’ve been privileged to have Alekh visiting me in NY where he will be sorely missed.
- John Duchi created Adagrad which is a commonly helpful improvement over online gradient descent that is seeing wide adoption, including in Vowpal Wabbit. He has a similarly deep and broad understanding of optimization and learning with significant industry experience at Google. Alekh and John have often coauthored together.
- Stephane Ross visited me a year ago over the summer, implementing many new algorithms and working out the first scale free online update rule which is now the default in Vowpal Wabbit. Stephane is not on the market—Google robbed the cradle successfully I’m sure that he will do great things.
- Anna Choromanska visited me this summer, where we worked on extreme multiclass classification. She is very good at focusing on a problem and grinding it into submission both in theory and in practice—I can see why she wins awards for her work. Anna’s future in research is quite promising.
I also wanted to mention some postdoc openings in machine learning.
- In New York Leon Bottou, Miro Dudik, and I are looking for someone. The deadline is December 13.
- In New England, Sham Kakade and Adam Kalai are looking for someone. The deadline is December 13.
- Also in the New York area, Daniel Hsu and Tong Zhang may both be considering a postdoc with no particular deadline.
- In England, Peter Flach is looking for two postdocs on a health & machine learning project with a deadline of December 2. I consider machine learning for healthcare of critical importance in the future.
There will be no New York ML Symposium this year. The core issue is that NYAS is disorganized by people leaving, pushing back the date, with the current candidate a spring symposium on March 28. Gunnar and I were outvoted here—we were gung ho on organizing a fall symposium, but the rest of the committee wants to wait.
In some good news, most of the ICML 2012 videos have been restored from a deep backup.
Manik and I are organizing the extreme classification workshop at NIPS this year. We have a number of good speakers lined up, but I would further encourage anyone working in the area to submit an abstract by October 9. I believe this is an idea whose time has now come.
The NIPS website doesn’t have other workshops listed yet, but I expect several others to be of significant interest.
A big ouch—all the videos for ICML 2012 were lost in a shuffle. Rajnish sends the below, but if anyone can help that would be greatly appreciated.
Sincere apologies to ICML community for loosing 2012 archived videos
What happened: In order to publish 2013 videos, we decided to move 2012 videos to another server. We have a weekly backup service from the provider but after removing the videos from the current server, when we tried to retrieve the 2012 videos from backup service, the backup did not work because of provider-specific requirements that we had ignored while removing the data from previous server.
What are we doing about this: At this point, we are still looking into raw footage to find if we can retrieve some of the videos, but following are the steps we are taking to make sure this does not happen again in future:
(1) We are going to create a channel on Vimeo (and potentially on YouTube) and we will publish there the p-in-p- or slide-versions of the videos. This will be available by the beginning of Oct 2013.
(2) We are going to provide download links from TechTalks so that the slide-version (of p-in-p- version if availbale) of the videos can be directly downloaded by viewers.This feature will be available by Aug 4th 2013.
(3) Of course we are now creating regular backups that do not depend on our service provider.
How can you help: If you have downloaded from TechTalks the ICML 2012 videos using external tools, we will really appreciate if you can provide us the videos, please email at email@example.com .
This post is a (near) transcript of a talk that I gave at the ICML 2013 Workshop on Peer Review and Publishing Models. Although there’s a PDF available on my website, I’ve chosen to post a slightly modified version here as well in order to better facilitate discussion.
Disclaimers and Context
I want to start with a couple of disclaimers and some context.
First, I want to point out that although I’ve read a lot about double-blind review, this isn’t my research area and the research discussed in this post is not my own. As a result, I probably can’t answer super detailed questions about these studies.
I also want to note that I’m not opposed to open peer review — I was a free and open source software developer for over ten years and I care a great deal about openness and transparency. Rather, my motivation in writing this post is simply to create awareness of and to initiate discussion about the benefits of double-blind review.
Lastly, and most importantly, I think it’s essential to acknowledge that there’s a lot of research on double-blind review out there. Not all of this research is in agreement, in part because it’s hard to control for all the variables involved and in part because most studies involve a single journal or discipline. And, because these studies arise from different disciplines, they can be difficult to
track down — to my knowledge at least, there’s no “Journal of Double-Blind Review Research.” These factors make for a hard landscape to navigate. My goal here is therefore to draw your attention to some of the key benefits of double-blind review so that we don’t lose sight of them when considering alternative reviewing models.
How Blind Is It?
The primary motivation behind double-blind peer review — in which the identities of a paper’s authors and reviewers are concealed from each other — is to eliminate bias in the reviewing process by preventing factors other than scientific quality from influencing the perceived merit of the work under review. At this point in time, double-blind review is the de facto standard for machine learning conferences.
Before I discuss the benefits of double-blind review, however, I’d like to address one of its most commonly heard criticisms: “But it’s possible to infer author identity from content!” — i.e., that double-blind review isn’t really blind, so therefore there’s no point in implementing it. It turns out that there’s some truth to this statement, but there’s also a lot of untruth too. There are several studies that directly test this assertion by asking reviewers whether authors or institutions are identifiable and, if so, to record their identities and describe the clues that led to their identification.
The results are pretty interesting: when asked to guess the identities of authors or institutions, reviewers are correct only 25–42% of the time . The most common identification clues are self-referencing and authors’ initials or institution identities in the manuscript, followed by reviewers’ personal knowledge [2, 3]. Furthermore, higher identification percentages correspond to journals in which papers are required to explicitly state the source of the data being studied . This indicates that journals, not just authors, bear some responsibility for the degree of identification clues present and can therefore influence the extent to which review is truly double-blind.
Is It Necessary?
Another commonly heard criticism of double-blind review is “But I’m not biased!” — i.e., that double-blind review isn’t needed because factors other than scientific quality do not affect reviewers’ opinions anyway. It’s this statement that I’ll mostly be focusing on here. There are many studies that address this assertion by testing the extent to which peer review can be biased against new ideas, women, junior researchers, and researchers from less prestigious universities or countries other than the US. In the remainder of this post, I’m therefore going give a brief overview of these studies’ findings. But before I do that, I want to talk a bit more about bias.
I think it’s important to talk about bias because I want to make it very clear that the kind of bias I’m talking about is NOT necessarily ill-intentioned, explicit, or even conscious. To quote the AAUW’s report  on the under-representation of women in science, “Even individuals who consciously refute gender and science stereotypes can still hold that belief at an unconscious level. These unconscious beliefs or implicit biases may be more powerful than explicitly held beliefs and values simply because we are not aware of them.” Chapters 8 and 9 of this report provide a really great overview of recent research on implicit bias and negative stereotypes in the workplace. I highly recommend reading them — and the rest of the report for that matter — but for the purpose of this post, it’s sufficient to remember that “Less-conscious beliefs underlying negative stereotypes continue to influence assumptions about people and behavior. [Even] good people end up unintentionally making decisions that violate [...] their own sense of what’s correct [and] what’s good.”
Prestige and Familiarity
Perhaps the most well studied form of bias is the “Matthew effect,” originally introduced by Robert Merton in 1968 . This term refers to the “rich-get-richer” phenomenon whereby well known, eminent researchers get more credit for their contributions than unknown researchers. Since 1968, there’s been a considerable amount of follow-on research investigating the extent to which the Matthew effect exists in science. In the context of peer review, reviewers may be more likely to recommend acceptance of incomplete or inferior papers if they are authored by more prestigious researchers.
Country of Origin
It’s also important to consider country of origin and international bias. There’s research  showing that reviewers from within the United States and reviewers from outside the United States evaluate US papers more favorably, with US reviewers showing a stronger preference for US papers than non-US reviewers. In contrast, US and non-US reviewers behaved near identically for non-US papers.
One of the most widely discussed pieces of recent work on double-blind review and gender is that of Budden et al. , whose research demonstrated that following the introduction of double-blind review by the journal Behavioral Ecology, there was a significant increase in papers authored by women. This pattern was not observed in a similar journal that instead reveals author information to reviewers. Although there’s been some controversy surrounding this work , mostly questioning whether the observed increase was indeed to do with the policy change or a more widely observed phenomenon, the original authors reanalyzed their data and again found that double-blind review favors increased representation of female authors .
Race has also been demonstrated to influence reviewers’ recommendations, albeit in the context of grant funding rather than publications. Even after controlling for factors such as educational background, country of origin, training, previous research awards, publication record, and employer characteristics, African-American applicants for National Institutes of Health R01 grants are 10% less likely than white applicants to be awarded research funding .
I also want to talk briefly about stereotype threat. Stereotype threat is a phenomenon in which performance in academic contexts can be harmed by the awareness that one’s behavior might be viewed through the lens of a negative stereotype about one’s social group . For example, studies have demonstrated that African-American students enrolled in college and female students enrolled in math and science courses score much lower on tests when they are reminded beforehand of their race or gender [10, 11]. In the case of female science students, simply having a larger ratio of men to women present in the testing situation can lower women’s test scores . Several factors may contribute to this decreased performance, including the anxiety, reduced attention, and self-consciousness associated with worrying about whether or not one is confirming the stereotype. One idea that that hasn’t yet been explored in the context of peer review, but might be worth investigating, is whether requiring authors to reveal their identities during peer review induces a stereotype threat scenario.
Lastly, I want to mention the identification of reviewers. Although there’s much less research on this side of the equation, it’s definitely worth considering the effects of revealing reviewer identities as well — especially for more junior reviewers. To quote Mainguy et al.’s article  in PLoS Biology, “Reviewers, and especially newcomers, may feel pressured into accepting a mediocre paper from a more established lab in fear of future reprisals.”
I want to conclude by reminding you that my goal in writing this post was to create awareness about the benefits of double-blind review. There’s a great deal of research on double-blind review and although it can be a hard landscape to navigate — in part because there are many factors involved, not all of which can be trivially controlled in experimental conditions — there are studies out there that demonstrate concrete benefits of double-blind review. Perhaps more importantly though, double-blind review promotes the PERCEPTION of fairness. To again quote Mainguy et al., “[Double-blind review] bears symbolic power that will go a long way to quell fears and frustrations, thereby generating a better perception of fairness and equality in global scientific funding and publishing.”
 Budden, Tregenza, Aarssen, Koricheva, Leimu, Lortie. “Double-blind review favours increased representation of female authors.” 2008.
 Yankauer. “How blind is blind review?” 1991.
 Katz, Proto, Olmsted. “Incidence and nature of unblinding by authors: our experience at two radiology journals with double-blinded peer review policies.” 2002.
 Hill, Corbett, St, Rose. “Why so few? Women in science, technology, engineering, and mathematics.” 2010.
 Merton. “The Matthew effect in science.” 1968.
 Link. “US and non-US submissions: an analysis of reviewer bias.” 1998.
 Webb, O’Hara, Freckleton. “Does double-blind review benefit female authors?” 2008.
 Budden, Lortie, Tregenza, Aarssen, Koricheva, Leimu. “Response to Webb et al.: Double-blind review: accept with minor revisions.” 2008.
 Ginther, Schaffer, Schnell, Masimore, Liu, Haak, Kington. “Race, ethnicity, and NIH research awards.” 2011.
 Steele, Aronson. “Stereotype threat and the intellectual test performance of African Americans.” 1995.
 Dar-Nimrod, Heine. “Exposure to scientific theories affects women’s math performance.” 2006,
 Mainguy, Motamedi, Mietchen. “Peer review—the newcomers’ perspective.” 2005.
When thinking about how best to review papers, it seems helpful to have some conception of what good reviewing is. As far as I can tell, this is almost always only discussed in the specific context of a paper (i.e. your rejected paper), or at most an area (i.e. what a “good paper” looks like for that area) rather than general principles. Neither individual papers or areas are sufficiently general for a large conference—every paper differs in the details, and what if you want to build a new area and/or cross areas?
An unavoidable reason for reviewing is that the community of research is too large. In particular, it is not possible for a researcher to read every paper which someone thinks might be of interest. This reason for reviewing exists independent of constraints on rooms or scheduling formats of individual conferences. Indeed, history suggests that physical constraints are relatively meaningless over the long term — growing conferences simply use more rooms and/or change formats to accommodate the growth.
This suggests that a generic test for paper acceptance should be “Are there a significant number of people who will be interested?” This question could theoretically be answered by sending the paper to every person who might be interested and simply asking them. In practice, this would be an intractable use of people’s time: We must query far fewer people and achieve an approximate answer to this question. Our goal then should be minimizing the approximation error for some fixed amount of reviewing work.
Viewed from this perspective, the first way that things can go wrong is by misassignment of reviewers to papers, for which there are two
easy failure modes available.
- When reviewer/paper assignment is automated based on an affinity graph, the affinity graph may be low quality or the constraint on the maximum number of papers per reviewer can easily leave some papers with low affinity to all reviewers orphaned.
- When reviewer/paper assignments are done by one person, that person may choose reviewers who are all like-minded, simply because
this is the crowd that they know. I’ve seen this happen at the beginning of the reviewing process, but the more insidious case is when it happens at the end, where people are pressed for time and low quality judgements can become common.
An interesting approach for addressing the constraint objective would be optimizing a different objective, such as the product of affinities
rather than the sum. I’ve seen no experimentation of this sort.
For ICML, there are about 3 levels of “reviewer”: the program chair who is responsible for all papers, the area chair who is responsible for organizing reviewing on a subset of papers, and the program committee member/reviewer who has primary responsibility for reviewing. In 2012 tried to avoid these failure modes in a least-system effort way using a blended approach. We used bidding to get a higher quality affinity matrix. We used a constraint system to assign the first reviewer to each paper and two area chairs to each paper. Then, we asked each area chair to find one reviewer for each paper. This obviously dealt with the one-area-chair failure mode. It also helps substantially with low quality assignments from the constrained system since (a) the first reviewer chosen is typically higher quality than the last due to it being the least constrained (b) misassignments to area chairs are diagnosed at the beginning of the process by ACs trying to find reviewers (c) ACs can reach outside of the initial program committee to find reviewers, which existing automated systems can not do.
The next way that reviewing can go wrong is via biased reviewing.
- Author name bias is a famous one. In my experience it is real: well known authors automatically have their paper taken seriously, which particularly matters when time is short. Furthermore, I’ve seen instances where well-known authors can slide by with proof sketches that no one fully understands.
- Review anchoring is a very significant problem if it occurs. This does not happen in the standard review process, because the reviews of others are not visible to other reviewers until they are complete.
- A more subtle form of bias is when one reviewer is simply much louder or charismatic than others. Reviewing without an in-person meeting is actually helpful here, as it reduces this problem substantially.
Reviewing can also be low quality. A primary issue here is time: most reviewers will submit a review within a time constraint, but it may not be high quality due to limits on time. Minimizing average reviewer load is quite important here. Staggered deadlines for reviews are almost certainly also helpful. A more subtle thing is discouraging low quality submissions. My favored approach here is to publish all submissions nonanonymously after some initial period of time.
Another significant issue in reviewer quality is motivation. Making reviewers not anonymous to each other helps with motivation as poor reviews will at least be known to some. Author feedback also helps with motivation, as reviewers know that authors will be able to point out poor reviewing. It is easy to imagine that further improvements in reviewer motivation would be helpful.
A third form of low quality review is based on miscommunication. Maybe there is silly typo in a paper? Maybe something was confusing? Being able to communicate with the author can greatly reduce ambiguities.
The last problem is dictatorship at decision time for which I’ve seen several variants. Sometimes this comes in the form of giving each area chair a budget of papers to “champion”. Sometimes this comes in the form of an area chair deciding to override all reviews and either accept or more likely reject a paper. Sometimes this comes in the form of a program chair doing this as well. The power of dictatorship is often available, but it should not be used: the wiser course is keeping things representative.
At ICML 2012, we tried to deal with this via a defined power approach. When reviewers agreed on the accept/reject decision, that was the decision. If the reviewers disgreed, we asked the two area chairs to make decisions and if they agreed, that was the decision. It was only when the ACs disagreed that the program chairs would become involved in the decision.
The above provides an understanding of how to create a good reviewing process for a large conference. With this in mind, we can consider various proposals at the peer review workshop and elsewhere.
- Double Blind Review. This reduces bias, at the cost of decreasing reviewer motivation. Overall, I think it’s a significant long term positive for a conference as “insiders” naturally become more concerned with review quality and “outsiders” are more prone to submit.
- Better paper/reviewer matching. A pure win, with the only caveat that you should be familiar with failure modes and watch out for them.
- Author feedback. This improves review quality by placing a check on unfair reviews and reducing miscommunication at some cost in time.
- Allowing an appendix or ancillary materials. This allows authors to better communicate complex ideas, at the potential cost of reviewer time. A standard compromise is to make reading an appendix optional for reviewers.
- Open reviews. Open reviews means that people can learn from other reviews, and that authors can respond more naturally than in single round author feedback.
It’s important to note that none of the above are inherently contradictory. This is not necessarily obvious as proponents of open review and double blind review have found themselves in opposition at times. These approaches can be accommodated by simply hiding authors names for a fixed period of 2 months while the initial review process is ongoing.
Representative reviewing seems like the real difficult goal. If a paper is rejected in a representative reviewing process, then perhaps it is just not of sufficient interest. Similarly, if a paper is accepted, then perhaps it is of real and meaningful interest. And if the reviewing process is not representative, then perhaps we should fix the failure modes.
Edit: Crossposted on CACM.
The large scale machine learning class I taught with Yann LeCun has finished. As I expected, it took quite a bit of time :-). We had about 25 people attending in person on average and 400 regularly watching the recorded lectures which is substantially more sustained interest than I expected for an advanced ML class. We also had some fun with class projects—I’m hopeful that several will eventually turn into papers.
I expect there are a number of professors interested in lecturing on this and related topics. Everyone will have their personal taste in subjects of course, but hopefully there will be some convergence to common course materials as well. To help with this, I am making the sources to my presentations available. Feel free to use/improve/embelish/ridicule/etc… in the pursuit of the perfect course.
ICML registration is also available, at about an x3 higher cost. My understanding is that this is partly due to the costs of a larger conference being harder to contain, partly due to ICML lasting twice as long with tutorials and workshops, and partly because the conference organizers were a bit over-conservative in various ways.
Adam Kalai points out the New England Machine Learning Day May 1 at MSR New England. There is a poster session with abstracts due April 19. I understand last year’s NEML went well and it’s great to meet your neighbors at regional workshops like this.
Sebastien Bubeck has a new ML blog focused on optimization and partial feedback which may interest people.
Yann and I have arranged so that people who are interested in our large scale machine learning class and not able to attend in person can follow along via two methods.
- Videos will be posted with about a 1 day delay on techtalks. This is a side-by-side capture of video+slides from Weyond.
- We are experimenting with Piazza as a discussion forum. Anyone is welcome to subscribe to Piazza and ask questions there, where I will be monitoring things. update2: Sign up here.
The first lecture is up now, including the revised version of the slides which fixes a few typos and rounds out references.