The Erdős number, rebuilt for the deep-learning era. Paul Erdős wrote 1,500+ papers; mathematicians measure their distance to him in co-authorship hops. We do the same with Geoffrey E. Hinton as the origin.
Co-author a paper with Hinton and your number is 1; co-author with a 1 and yours is 2 — your Hinton number is the length of the shortest co-authorship path from you back to Hinton.
Everything comes from the full dblp.org dump (CC0 1.0) — the computer-science bibliography. We rebuild the co-author graph and re-run the computation monthly; the verified date on every badge is that build date. Your number and the path behind it are precomputed, not computed on request.
Of the 4.2M people in DBLP, 0% have a finite path to Hinton. Each square below is ~21,000 people — the gray ones have published, but never within reach of the giant connected component.
Only 365 people share a paper byline with Hinton — the top 0.009% of connected people. Not all of those are close collaborations; some share only a large consortium byline. Being a 1 is about as rare, among CS authors, as being struck by lightning in your lifetime.
The Hinton number is measured in a unit invented for Paul Erdős — and Erdős himself has a Hinton number of 3. So does everyone below: some helped invent computing, some spent a career arguing the field was on the wrong track, some never went near it. Publish anywhere near the field's ancestry and the graph finds you.
The most common distance is 4 — 53% of connected people sit there, and 96% land between 3 and 5. Distances above 7 are vanishingly rare.
Mean distance to the center of each famous small world. The average Hinton number beats the average Erdős number — deep learning collaborates harder.
Hop count treats a one-off workshop paper the same as a decade of collaboration. Collaboration strength doesn't. We use the standard co-authorship tie strength from network science — Newman's collaboration weight (M. E. J. Newman, 2001) — where each shared paper adds 1 / (co-authors − 1) to an edge: many small-team papers make a strong edge; a 450-author consortium paper adds almost nothing.
Among your shortest paths — all the same number of hops — we take the one whose co-authorship is tightest, and use its weight to place your square inside your own band on the grid on your page: everyone with your Hinton number, strongest chains first. A 3 is only ever ordered against other 3s. Hops stay primary — no amount of collaboration weight turns a 4 into a 3. One honest caveat: much of that ordering is inherited from which of Hinton's direct co-authors — the distance-1 author your best chain runs through — rather than from your own collaborations, so read it as a curiosity more than a verdict. For example:
Why is my number ∞? You have no path to Hinton in DBLP yet — usually meaning no indexed publication. Your first paper is your ticket in.
I can't find myself / that isn't me. Names collide; pick yourself by the paper titles shown in search. You can also type a venue or a word from one of your titles after your name — wei wang vldb — since search covers your two most recent papers too. Author identity follows DBLP's disambiguation.
Non-CS papers? Not yet — DBLP covers computer science. Broader sources are on the roadmap.