Dr. David Balduzzi

Adresse: Spemannstr. 38
72076 Tübingen
Raum Nummer: 222
Tel.: 07071 601 584
Fax: 07071 601 552
E-Mail: david.balduzzi


Bild von Balduzzi, David, Dr.

David Balduzzi

Position: Wissenschaftler  Abteilung: 

I work in mathematical neuroscience where I apply methods from information theory, machine learning and algebraic geometry. The main question I'm working on is: How do neurons learn to cooperate? Statistical learning theory has found necessary and sufficient conditions for the fast convergence of single classifiers to optimal solutions. I am using geometric methods to extend these results to distributed systems, taking inspiration from neurophysiology.


Please see my research statement or webpage for more details.

My project is to understand distributed learning. The paradigmatic example of a distributed learning system is the mammalian cortex, which contains about 100 billion learners (the individual neurons). I hope to discover the features of neuronal and cortical organization that are essential to distributed learning, and distill them into computational principles amenable to mathematical analysis and algorithmic implementation.
  • How do neurons learn to cooperate?

The human brain contains about 100,000,000,000 neurons and 1,000,000,000,000,000 synapses. Most neurons do not interact directly with the environment but only with other neurons. All neurons "see" is patterns of activity on their 10,000 or so synapses, slices of about 0.00001% of total brain activity. All neurons do is output sequences of spikes through their single axons (each connecting to 10,000 or so other neurons), whilst having no idea how billions of other neurons reinterpret their outputs. Individual neurons are ignorant and ineffective. They do not know what their inputs mean, they do not know what they are communicating, to whom, or for what purpose, and in any case they have very little to say. How, then, can neurons possibly act in the interest of their brain?

  • What kind of language should inductive reasoners use to communicate?

The basis of scientific knowledge is induction: generalizing from observations. Vapnik and Chervonenkis proved necessary and sufficient conditions for consistency and fast convergence of induction via empirical risk minimization by suitably quantifying the complexity of learning algorithms, thereby laying the foundation for machine learning. However, there is more to scientific -- objective -- knowledge than generalizing from observations. According to Karl Popper, "a justification is objective if in principle it can be tested and understood by anybody". Thus, knowledge should be communicable and reproducible to count as objective. Whilst deduction and induction have been adequately formalized (Turing machines and statistical learning theory), communicability has not. The brain can be considered as a vast collection of inductive reasoners (the individual neurons) that communicate what they learn via spikes. This suggests that the neural code may be a highly effective language for communicating generalizations.

Referenzen pro Seite: Jahr: Medium:

Zeige Zusammenfassung

Artikel (6):

Balduzzi D, Ortega PA und Besserve M (Mai-2013) Metabolic cost as an organizing principle for cooperative learning Advances in Complex Systems 16(02n03) 1-18.
Balduzzi D und Tononi G (August-2009) Qualia: The Geometry of Integrated Information PLoS Computational Biology 5(8) 1-24.
Balduzzi D, Riedner BA und Tononi G (Oktober-2008) A BOLD window into brain waves Proceedings of the National Academy of Sciences of the United States of America 105(41) 15641-15642.
Balduzzi D und Tononi G (Juni-2008) Integrated Information in Discrete Dynamical Systems: Motivation and Theoretical Framework PLoS Computational Biology 4(6) 1-18.
Balduzzi D (März-2008) Poisson Geometry of Parabolic Bundles on Elliptic Curves International Journal of Mathematics 19(3) 339-367.
Balduzzi D (November-2006) Donagi-Markman cubic for Hitchin systems Mathematical Research Letters 13(6) 923-933.

Beiträge zu Tagungsbänden (8):

Ortega PA, Grau-Moya J, Genewein T, Balduzzi D und Braun DA (April-2013) A Nonparametric Conjugate Prior Distribution for the Maximizing Argument of a Noisy Function In: Advances in Neural Information Processing Systems 25, , Twenty-Sixth Annual Conference on Neural Information Processing Systems (NIPS 2012), Curran, Red Hook, NY, USA, 3014-3022.
Balduzzi D und Besserve M (April-2013) Towards a learning-theoretic analysis of spike-timing dependent plasticity In: Advances in Neural Information Processing Systems 25, , Twenty-Sixth Annual Conference on Neural Information Processing Systems (NIPS 2012), Curran, Red Hook, NY, USA, 2465-2473.
Balduzzi D (2013) Falsification and future performance In: Algorithmic Probability and Friends: Bayesian Prediction and Artificial Intelligence, , Ray Solomonoff 85th Memorial Conference 2011, Springer, Berlin, Germany, 65-78, Series: Lecture Notes in Computer Science ; 7070.
Balduzzi D (Dezember-2011) Information, learning and falsification, NIPS 2011 Philosophy and Machine Learning Workshop, 1-4.
Balduzzi D (September-2011) Estimating integrated information with TMS pulses during wakefulness, sleep and under anesthesia, 33rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (IEEE EMBC 2011), IEEE, Piscataway, NJ, USA, 4717-4720.
Balduzzi D (August-2011) Detecting emergent processes in cellular automata with excess information In: Advances in Artificial Life: ECAL 2011, , Eleventh European Conference on the Synthesis and Simulation of Living Systems, MIT Press, Cambridge, MA, USA, 55-62.
Balduzzi D (Juli-2011) On the information-theoretic structure of distributed measurements, 7th International Workshop on Developments of Computational Models (DCM 2011), Elsevier Science, Amsterdam, Netherlands, 1-15.
Gomez Rodriguez M, Balduzzi D und Schölkopf B (Juli-2011) Uncovering the Temporal Dynamics of Diffusion Networks, 28th International Conference on Machine Learning (ICML 2011), International Machine Learning Society, Madison, WI, USA, 561-568.

Beiträge zu Büchern (1):

Tononi G und Balduzzi D: Toward a Theory of Consciousness, 1201-1220. In: The Cognitive Neurosciences, (Ed) M.S. Gazzaniga, MIT Press, Cambridge, MA, USA, (Oktober-2009).

Export als:
BibTeX, XML, pubman, Edoc, RTF
Last updated: Montag, 22.05.2017