R.C. Mintram and J. Vincent (UK)
Connectionist models, vector representa tion (VREP), General Encoder Decoder (GED), recovery profile (VRP)
A consequence of the connectionist approach to artificial intelligence is the requirement for structured data to be encoded into fixed width vector representations (VREP). This paper describes a graphical representation of the width of a VREP plotted against the degree to which the data has been successfully encoded. This is referred to as a VREP Recovery Profile (VRP). The VRP enables the optimum vector width to be identified, and this may not be that automatically determined by, for example, an (S)RAAM. Additionally, it allows some characteristics of the training/generalisation to be examined. The recent development of a General Encoder Decoder framework (GED) allows the vector width to be varied and supports the use of the VRP.
Important Links:
Go Back
