====== Jürgen Schmidhuber ======
{{:coursework:2014f:juergen.gif?200 |}}

Jürgen seems to be a god-tier researcher in the fields of neural networks, machine learning and artificial intelligence. He and his research teams have won nine international competitions in machine learning and pattern recognition. He has an incredible list of academic accolades that seem to fly his way whenever he touches something.

Diploma: 1987 from Technische Universität München  (4 years compared to an average of 6.05)

PhD: 1991 Technische Universität München //Dynamic Neural Nets and the Fundamental Spatio-Temporal Credit Assignment Problem//

Tenure awarded in 1993 at age 30 (German average is 40 years old)

His [[http://people.idsia.ch/~juergen/|homepage]]

===== Initial Impression =====
He cannot be human. He is an absolute monster when it comes to cutting edge research in A-Life.

  * Number 1458 in the most cited computer science authors (2008).
  * Established the field of Universal AI (universal way of predicting)
  * He generalized algorithmic information theory, and the many-worlds theory of physics, to obtain a minimal theory of all constructively computable universes
  * Essentially started the field of Deep Learning by writing the first 'Long Short-Term Memory' Recurrent Neural Network (Google, IBM & Microsoft now employ these techniques)
  * An incredibly amount of firsts (first deep learners to win pattern recognition contests in general, first to win object detection contests, first to win pure image segmentation contest, first superhuman visual recognition performance)
  * Wrote the first genetic programming system with loops and variable length code -- published the second paper on GP


===== Field Contributions =====
He has made a myriad of contributions to the following fields:

  * Neural Networks (with LSTM RNNs) 
  * Deep Learning & Computer Vision (Fast Deep Neural Nets, Learning attentive vision - identifying and tracking moving targets)
  * Artificial Evolution (Yay! Genetic programming, Meta GP, Probabilistic incremental program evolution)
  * Reinforcement Learning
  * Unsupervised Learning
  * Metalearning Machines
  * Swarm Intelligence (artificial ants, breaking records and benchmarks)
  * Optimally efficient, universal problem solver (Gödel machine - provably correct, more efficient rewrites of it's own code and functions or algorithms it is applied to)
    * Optimal Ordered Problem Solver
  * Universal Learning Algorithms ("There is a theoretically optimal way of predicting the future, given the past. It can be used to define an optimal (though noncomputable) rational agent...")
  * Algorithmic Theories of Everything (...)
    * Analyzes all the universes with limit-computable probabilities as well as the very limits of formal describability. 
  * Super Omegas and Generalized Kolmogorov Complexity and Algorithmic Probability (???)
  * Made a formal theory of creativity... wat ({{:coursework:2014f:femmefractale466.gif?linkonly|Low-Complexity Art)}})