Plague of dimension

Plague of dimension or Malédiction of dimension ( Curse off dimensionality ) is a term invented by Richard Bellman to qualify the problem of the explosive increase in volumes of data associated with the addition of additional size in a mathematical space.

Leo Breiman gives the example of 100 observations covering the unidimensional interval in realities: it is possible to draw up a Histogramme results and to draw some from the Inférence S. On the other hand, in space corresponding to 10 dimensions 10, the 100 observations are points isolated in a vast empty space, and do not allow the statistical analysis. To carry out in 10 a cover equivalent to that of the 100 point in, one does not need less 1020 observations - gigantic and often impracticable company.

The plague of dimension is a major hurdle in the machine learning, which often amounts drawing from the inférences of a reduced number of experiments in a space of possibilities of high size. It then becomes often necessary to inject information A priori so as to force the system of training to obtain inférences. It must be prepared with the type of information to extract. One speaks then about Inférence bayésienne.

Random links:Special organization | RS Gimnástica de Torrelavega | Low-Cerdagne | Timothy Cox | Turn of Georgia | Mickleham,_Victoria