We used L2 penalization (penalizing the squared norm of the parameter vector).
To fit the models in this situation, one would need to reduce the number of parameters that one tries to estimate, either manually, or perhaps by use of some kind of penalization approach [ Li et al., 2009].
Some of the most common penalization methods are the least absolute shrinkage and selection operator (lasso) [ 6], smoothly clipped absolute deviation (SCAD) [ 7], Dantzig selector [ 8], LARS [ 9], and the smooth integration of counting and absolute deviation (SICA) penalty [ 10].
In our experiments, we show that in most tests DPM outperforms both Pregel and Fork-Join in terms of recommendation time, with a minor penalization in network usage in some scenarios.
Also, penalization techniques may suffer from some drawbacks.
According to the ill-posed behavior of the inverse problem, some regularization tools must be performed and the Tikhonov penalization type is the most commonly used in optical tomography applications.
In the case of the multiplicative model, the penalization can reach 100%, a point that to some extent contradicts our deep-seated survival instinct.
These approaches introduce l1 (Lasso) penalization terms on the loading vectors to shrink some of the coefficients towards zero, thus allowing for simultaneous variables selection in the two data sets.
All penalized regression methods require input of one or more values for the penalization parameter(s).
