in Manno (Switzerland), since January 2003.
As of 2013: A dozen previous
PhD students and postdocs of JS became professors at various universities.
Recipient of the 2013 Helmholtz Award of
the International Neural Networks Society
2009: Appointed to the
Academia Scientiarum et Artium Europaea, the
European Academy of Sciences and Arts, whose 1317 members include numerous Nobel laureates, renowned
architects and artists, and the Pope.
AI Lab Rankings
2 years after JS became director
the 1997 "X-Lab Survey" by the US journal Business Week
Magazine ranked IDSIA among the world's top ten AI labs.
(The other members of the top 10 club were much larger than IDSIA.)
IDSIA also was ranked 4th in the category "biologically inspired."
(IDSIA's most important results came later though.)
1st Prizes in International Competitions
The deep/recurrent neural networks (Deep Learning systems) developed in the lab of JS won nine international pattern recognition contests:
9. MICCAI 2013 Grand Challenge on Mitosis Detection
8. ICPR 2012 Contest on Mitosis Detection in Breast Cancer Histological Images (MITOS)
6. IJCNN 2011 on-site Traffic Sign Recognition Competition (0.56% error rate, the only method outperforming humans, August 2011)
5. ICDAR 2011 Chinese Handwriting Recognition Competition (June 2011)
4. Online German Traffic Sign Recognition Contest (1st & 2nd rank; 1.02% error rate, January 2011)
3. ICDAR 2009 Arabic Connected Handwriting Competition (like the competitions below
won by multi-dimensional LSTM recurrent neural networks trained by CTC)
2. ICDAR 2009 Handwritten Farsi/Arabic Character Recognition Competition
1. ICDAR 2009 French Connected Handwriting Competition
Best Papers etc
JS thinks his best papers were published in journals or at
conferences without best paper awards.
Nevertheless, he and his co-workers got best paper awards at (1) GECCO 2005, (2) GECCO 2009, (3) AGI 2010 (Kurzweil Prize), (4) AGI 2011 (Kurzweil Prize), (5) AGI 2013, as well as additional best paper nominations at GECCO 2005, IROS 2006, GECCO 2009, GECCO 2010, PPSN 2012. JS' team also got the AAAI 2013 Best Student Video Award
1st Ranks in Important Machine Learning & Pattern Recognition Benchmarks
Methods developed in the lab of JS were the first to achieve human-competitive performance on important pattern recognition benchmarks, and set many records, e.g.:
5. Best recognition of Chinese characters from the ICDAR 2013 competition (3755 classes), approaching human performance (Sept 2013)
4. MNIST handwritten digits data set (perhaps the most famous machine
learning benchmark). New record (0.35% error rate) in 2010, broken
in June 2011 (0.27%), finally: human-competitive performance in Dec 2011 (0.23%)
for the first time in the history of this iconic benchmark.
3. NORB Object Recognition Benchmark, NYU, 2004. New records in 2011.
2. CIFAR-10 Object Recognition Benchmark, U. Toronto, 2009. New records in 2011.
1. Human Action Recognition Benchmarks, Weizmann Inst. and KTH. New records in 2010.
Numerous invited keynotes / plenary talks at
e. g., CIG 2013 keynote,
TEDx UHasselt 2012,
TEDx Lausanne 2012, Bionetics 2012 keynote,
IScIDE 2012 keynote (Nanjing, China),
IJCNN 2011 (San Jose, California),
World Science Festival 2011 (New York City),
Oxford Winter Conference on Intelligence 2011,
Joint Conferences ECML 2010 / PKDD 2010 (Barcelona; banquet talk),
GPTP 2010 (Ann Arbor, 25th anniversary of GP), Singularity Summit 2009 (New York City),
AGI 2009 (Washington), ICANN 2008 (Prague), Artificial Cognitive Systems 2008 (Munich),
KES 2008 (Zagreb),
ALT 2007 & DS 2007 (Sendai, Japan; the only joint invited lecture),
ACAT 2007 (Amsterdam), Art Meets Science 2007 (Munich),
Zuse Symposium 2006 (Berlin), GWAL 2006, Turing Days 2006 (Istanbul), ICANN 2005 (Warsaw),
NN 2005 (Porto), Data Ecologies 2005 (Linz), ANNIE 2004 (St. Louis), NN 2004 (Porto), AAMAS 2004 (Leeds),
and others. Many additional invited talks at
leading universities and institutes,
e. g., Caltech, Stanford,
Cambridge, Gatsby, King's College, Hongkong, Singapore, Kyoto, Tokyo,
ETHZ, EPFL, Princeton, MIT, Harvard, etc.
Until 1994: Involved in competitive research grants worth
several million DM/CHF/EUR/US$. Sources: Industry, DFG, BMFT.
1995-2008: Direct competitive funding for JS
at IDSIA (without funding for project partners): CHF 3'358'433 + EUR 284'300.
Sources: SNF, HASLER, CSEM, EU.
At TUM, JS also was an original PI of the Excellence Cluster
Cognitive Technical Systems. Total funding: > US$ 8m/year for 5 years through the Excellence
Initiative (> US$ 2.5 billion for 2006-2011),
which selected Munich's TUM and LMU as two of the three first
German "Elite Universities". (At TUM, JS also got an extra EUR 334'805 from the EU.)
Since 2009: JS' direct external funding for research projects starting/acquired at the Swiss AI Lab IDSIA (without funding for project partners): EUR 4'073'078 + CHF 1'848'366 = CHF 7'407'773 (based on exchange rates as of project start dates) = US$ 7.9m (as of 31.12.2012). Sources include EU, SNF, Industry.
Total direct competitive funding for academic projects up to 2012 (without funding for project partners): > US$ 12m.
of editorial boards of various journals
of numerous conferences
for all major journals and conferences in the field
Regular reviewer of proposals to
EU, NSF (US), SNF (Swiss), NWO (Dutch),
DFG (German), Humboldt (German),
and other international grant providers
Lectures at TUM, UC Boulder, Univ. Lugano,
in German or English,
on artificial intelligence,
bio-inspired optimization and machine learning
(introductory and advanced),
mathematically optimal general problem solvers,
algorithmic information theory and its applications.
Practical machine learning lab courses
and seminars on numerous AI topics
such as computer vision, evolutionary algorithms,
statistical learning theory, music & computer science,
game theory, and others.
Previous Positions until 1995
at TUM (somewhat comparable to an assistant professor in the US):
October 1993 - February 1995. Privatdozent since
at University of Colorado at Boulder (CU):
July 1991 - December 1992. (Declined postdoc
offer by Caltech.)
at TUM: March 1991 - June 1991
in a printery: 1982 - 1983
Military Service -
Obligatory service in German Army 1981 - 1982.
August 1993: qualification for tenure professorship in
computer science (at age 30, German average close to 40),
Technische Universität München (TUM).
Postdoctoral thesis: ``Net Architectures,
Objective Functions, and Chain Rule''. Referees:
Profs C. v. d. Malsburg, W. Brauer (supervisor), B. Radig.
preparation of ``very good''
PhD thesis on ``Dynamic Neural Nets and the
Fundamental Spatio-Temporal Credit Assignment Problem'' at
TUM. Defense: January 1991. Referees: Profs W. Brauer (supervisor),
1983 - 1987:
4 year study of computer science and mathematics at
TUM (average is 6.05 years). ``Very good'' diploma degree.
This top 10,000 citeseer list of 26 February 2008
(selected from roughly 1 million computer science authors)
finds over 2,000 references to the work of JS, but such lists
are highly questionable and obviously cannot tell us much about true impact: computer science pioneer
himself has less than 1,500 cites there - many scientists with comparatively negligible impact rank higher than him, and the superheroes of theory (Kurt Goedel) and hardware (Konrad Zuse) are not even on the list.
(To put things in perspective:
the highest-cited single author paper ever - by Swiss biologist U. K. Laemmli -
has a record 200,000 citations
per co-author.) JS wrote in Citation bubble about to burst? (Nature, vol. 469, p. 34, 2011): "Like the less-than-worthless collateralized debt obligations that drove the recent financial bubble, and unlike concrete goods and real exports, citations are easy to print and inflate.
Financial deregulation led to short-term incentives for bankers and rating agencies to overvalue their CDOs, bringing down entire economies. Likewise, today's academic rankings provide an incentive for professors to maximize citation counts instead of scientific progress [...] We may already be in the middle of a citation bubble - witness how relatively unknown scientists can now collect more citations than the most influential founders of their fields [...] Note that I write from the country with the most citations per capita and per scientist."
JS' own contributions to the citation bubble are visible in his Google Scholar Citation Profile.
fluent (mother tongue)
corrupted by Italian, but
can make slow conversation
can order pizza. Currently improving.
17.1.1963 in München.
German; Swiss permit C.
Univ. Prof. USI
Prof. SUPSI Dr. rer. nat. TUM Dr. rer. nat. habil. TUM Dipl. Inf. Univ. TUM
Ulrike Krommer (since May 3, 1997)
Julia (since September 10, 1997), Leonie (since January 25, 2000)
low-complexity art (see publications),
theory of beauty,
science fiction written by scientists,
various physical activities.