Who organised Big Learning

Big Learning had organisers and an advising committee for each of its three NIPS editions. These were workshop roles attached to a particular year. The names below follow those groups for 2011, 2012 and 2013, from the two-day meeting in Sierra Nevada, Spain, to the subsequent Lake Tahoe workshops. A programme committee reviewed the submissions.

The series brought several kinds of research into one discussion: learning algorithms, parallel and distributed systems, data management and applications. The edition-specific groups accompanied changes in emphasis, from the wide algorithm-and-systems scope of 2011 and 2012 to the database conversation in 2013. The workshop calendar and the author format describe the process that led from each call to the presentations.

2011: algorithms, systems and tools at scale

The first edition was titled Big Learning: Algorithms, Systems, and Tools for Learning at Scale. The NeurIPS description of the 2011 workshop identified a need for a shared venue across machine learning, databases, distributed systems and programming languages. It placed system builders, algorithm researchers and end users in the same discussion about learning from terabytes or petabytes of data.

That scope included practical case studies, tools and parallel algorithms. The two-day programme gave invited talks, contributed talks, spotlights, posters and tutorials different roles within the meeting. Its organiser and advising-committee groups therefore belonged to an event with a broad subject range, rather than a single technical method. The 2011 edition page explains the aims and shape of the meeting; its papers and invited-talk pages identify the programme contributions.

Organisers

Joseph Gonzalez
Sameer Singh
Alice Zheng
Graham Taylor
James Bergstra
Misha Bilenko
Yucheng Low

Advising committee

Sugato Basu
Alexander J. Smola
Michael Franklin
Andrew McCallum
Yoshua Bengio
Carlos Guestrin
Michael Jordan

The 2011 workshop's named application domains included bioinformatics, astronomy, recommendation systems, social networks, computer vision, web search and online advertising. That range helps explain why the meeting needed a shared discussion across research areas. Some questions concerned the supply and organisation of data; others concerned demanding models or the hardware and programming methods used to execute them. A common venue could put those different questions beside one another without making a single application define the scope of the workshop.

2012: data, models, applications and systems

The 2012 title, Big Learning: Algorithms, Systems, and Tools, continued the connection between learning methods and their implementation. The NeurIPS description of the 2012 workshop grouped the interests into Big Data, Models & Algorithms, Applications of Big Learning, and Tools, Software & Systems. Data cleaning, streaming observations, summaries and interpretation appeared beside parallel algorithms and scalable storage.

The meeting took place on 8 December at Lake Tahoe. Its one-day programme included invited talks, contributed work and poster sessions. The roles listed for that edition were distinct from presentation roles: being an organiser or adviser did not identify a person's particular paper or talk. The 2012 edition page connects this subject range to the programme, and its accepted-paper list gives each contribution's title and authors.

Organisers

Sameer Singh
John Duchi
Yucheng Low
Joseph Gonzalez

Advising committee

Alexander J. Smola
Michael Franklin
Andrew McCallum
Carlos Guestrin
Michael Jordan
Alice Zheng
Graham Taylor
James Bergstra
Misha Bilenko

2013: algorithms and data management

The final edition used the title Big Learning: Advances in Algorithms and Data Management. The NeurIPS description of the 2013 workshop made the exchange between database researchers and machine-learning researchers its central aim. It included system properties and implementation concepts alongside parallel learning algorithms. The call concerned online and batch learning, multicore and distributed execution, theoretical analysis and practical implementation studies.

The 9 December programme at Lake Tahoe reflected that breadth through invited talks, contributed talks, posters and a tutorial. The 2013 edition page explains the data-management focus, while the schedule shows how the different presentation formats fitted into the day. The organiser and advising-committee groups below are the groups associated with that edition, rather than a description of the invited programme.

Organisers

Xinghao Pan
Haijie Gu
Sameer Singh
Yucheng Low
Joseph Gonzalez

Advising committee

Carlos Guestrin
Andrew McCallum
Michael Jordan
Michael Franklin
Alexander J. Smola
Misha Bilenko
Alice Zheng
Markus Weimer
Tyson Condie

Roles within a workshop

The role labels matter because a research meeting brings together several kinds of participation. Organiser lists identify the people attached to the meeting's organisation; advising-committee lists identify a separate named group. Paper-author lists and invited-speaker lists answer different questions about the programme. A name appearing in more than one setting therefore has to be understood with the edition and role attached, rather than as a general description of someone's work.

The review process was another part of the event. Peer evaluation concerned the submissions that entered the accepted programme, whereas invited talks and tutorials supplied additional parts of the day. The author guidelines describe the extended-abstract format, the treatment of earlier publication and the contributed-talk or poster presentation categories. Together with the organiser groups, those rules show the structure of a workshop without making the role lists stand in for its research content.

The subjects themselves are developed in the topic guides. What big learning means explains the shared scale question, and learning inside databases develops the relationship foregrounded in 2013. The programmes show what was presented; the topic guides supply background for understanding the technical vocabulary that appeared across the editions.