regarding leveraging seq2seq to pure numerical string mapping
wenouyang opened this issue · comments
If the problem is pure sequence to sequence mapping, where each sequence consists of multiple numerical values/floating number. Can this platform be used to solve the problem.
For instance, one training pair sample is like x = [0.001, -0.11, 200.001]
, and y =[0.00, 0.123, 0.00]
. Physically, y
can be considered as a functional mapping of x, but y does not have a time dependence relationship as x
. This is different with the data set that you are using.
The test case will be also a string like test = [-200, 0.0023, -120]
, and we would like to get the mapping string of this “test” string. In other words, the use case is more like pure sequence-to-sequence mapping without time factor. Can this code be used to solve this kind of use case?