Prediction and Automatic Task Decomposition

[Rafal Salustowicz]

      LEARNING TO PREDICT THROUGH PROBABILISTIC INCREMENTAL
      PROGRAM  EVOLUTION  AND  AUTOMATIC TASK DECOMPOSITION

      Rafal Salustowicz                 Juergen Schmidhuber

                 Technical Report IDSIA-11-98

Analog gradient-based  recurrent  neural  nets  can learn complex
prediction tasks.  Most,   however,  tend to fail in case of long
minimal time lags between relevant training events.  On the other
hand, discrete methods such as search in a space of event-memori-
zing programs are not  necessarily  affected  at all by long time
lags:  we show that  discrete  "Probabilistic Incremental Program
Evolution" (PIPE) can solve several long time lag tasks that have
been successfully solved by only one analog method ("Long Short-
Term Memory" - LSTM).  In fact,  sometimes  PIPE even outperforms
LSTM. Existing discrete methods, however, cannot easily deal with
problems  whose solutions  exhibit comparatively high algorithmic
complexity.   We overcome this drawback by introducing filtering,
a novel,  general,  data-driven  divide-and-conquer technique for
automatic  task decomposition that is not limited to a particular
learning method. We compare PIPE plus filtering to various analog
recurrent net methods.

      ftp://ftp.idsia.ch/pub/rafal/TR-11-98-filter_pipe.ps.gz
      http://www.idsia.ch/~rafal/research.html

Rafal & Juergen, IDSIA, Switzerland                  www.idsia.ch