APRIL: Application of Probabilistic Inductive Logic Programming

1 January, 2004 to 31 December, 2006


  • Luc De Raedt

Note: This project was running at the ML lab in Freiburg, with Luc De Raedt as its principal investigator. This project addresses the problem of integrating probabilistic reasoning, first order logical representations and machine learning. This integration is one of the key open questions in artificial intelligence. An adequate answer to this open question is likely to result in new technologies that are applicable across a wide range of applications. Today, there is a thorough understanding of each of the three domains probability, logic, learning independently. There also exist some interesting results that combine two of these areas.

  • First, various techniques for probabilistic learning like gradient-based methods, the family of EM algorithms or Markov Chain Monte Carlo methods have been developed and exhaustively investigated in different communities, such as in the Uncertainty in AI community for Bayesian networks and in the Computational Linguistic community for Hidden Markov Models. These techniques are not only theoretically sound, they have also resulted in entirely new technologies for, and revolutionary novel products in computer vision, speech recognition, medical diagnostics, troubleshooting system (e.g. in Microsoft products), etc.
  • Second, Inductive Logic Programming has studied logic learning, i.e. learning and data mining within first order logic representations. Inductive Logic Programming has significantly broadened the application domain of data mining especially in bio- and chemoinformatics and now represent some of the best-known examples of Scientific Discovery by AI systems in the literature.
  • Third, researchers such as Nils Nilsson, Joseph Halpern and David Poole have also studied probabilistic logics.
  • Finally, there have been some initial attempts by e.g. the group around Daphne Koller in North-America or Taisuke Sato in Japan to approach the full problem of probabilistic logic learning. However, these last approaches are rather preliminary and not yet well-understood today.

Given the impact of both probabilistic learning and logic learning on technological developments in a wide range of applications, it is to be expected that effective probabilistic logic learning will pave the way towards an entirely new class of systems and applications. From a more scientific point of view, a better understanding of probabilistic logic learning is also likely to affect techniques for probabilistic learning and for learning logic. One can approach probabilistic logic learning from various sides. Given the long European tradition and experience in logic learning (i.e. inductive logic programming), we intend to approach the problem by extending Inductive Logic Programming techniques with probabilistic reasoning mechanisms. This also explains the title of the project, Application of Probabilistic Inductive Logic Programming.