You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Feb 12, 2024. It is now read-only.
The default paths are specified in a configuration file(see the training module for further details).
538
538
If the User specifes his own paths they have to be vaild.
539
539
Valid in this case means that the path points to a \gls{HDF5} file with the correct formatation.
540
-
If this is not the case the program will print an error message to the command line so that the user can specify a correct path.
540
+
If this is not the case the program will print an exception message to the command line so that the user can specify a correct path.
541
541
\subsubsection{Class Solver}
542
542
An Solver in our sense is an interative Solver which is able to to solve a linear system Ax=b for x, where A is a (in our case sparse) matrix of size nxn, x is a vector of size 1xn and b is vector of size 1xn(n$\in$N).
543
543
The matrix A in our case will be the Matrix the method execute(matrix,preconditioner) will receive, the vector b will be a random vector with values between 0 and 1.
This could be the case if the user interrupts the training process at a certain time and wants to to repeat the training later.
701
701
Other use cases are of course possible too.
702
702
The \gls{neural network} has to be a model of the Keras framework.
703
-
If the path is any other file the training module will print an error so that the user can specify a valid path.
703
+
If the path is any other file the training module will print an exception so that the user can specify a valid path.
704
704
If this path is not specified the training module will create a new \gls{neural network} (with the model definition and hyperparamters of the next category) and train with this one. \newline
705
705
706
706
The model definition and hyperparameters are used to determine which \gls{neural network} will be trained and tested.
The configuration file will determine from which path the labeled matrices will be loaded.
762
762
If there is no path specified in the method start, the default path will be used(see the class description of the configuration file).
763
763
The labeled matrices will be loaded in one \gls{HDF5} file.
764
-
If the path links to any other file, the class TrainingModule will print an error to the command line so that the user can specify a valid path. \newline
764
+
If the path links to any other file, the class TrainingModule will print an exception to the command line so that the user can specify a valid path. \newline
765
765
766
766
Then the values of the matrix will be normalized.
767
767
This means that every entry will be converted to a value between 0 and 1, such that relative distances are kept.
0 commit comments