Breve NN Abstract Data Type (ADT)
-
Nodes
-
int create-input-node with-name thisName =
"input-node" (string) with-bias thisBias = 0.0 (float)
with-initial-activation thisActivation =0.0 (float)
with-transfer-function thisTransferFunction = "Logistic" (string)
-
int create-hidden-node ...
-
int create-output-node ...
-
-
string get-node-name for-node thisNode (int).
-
int set-node-name for-node thisNode (int) to thisString (string).
-
get & set bias
-
get & set initial activation
-
get & set transfer function
-
get-current-activation
-
-
int delete-node thisNode (int).
-
Links
-
int create-link from-node thisNode (int) to-node thatNode (int) with-weight = 0.0 (float).
-
int set-link from-node thisNode (int) to-node thatNode (int) to-weight (float).
-
float get-link from-node thisNode (int) to-node thatNode (int).
-
int connection-exists from-node thisNode (int) to-node thatNode (int).
-
int delete-link from-node thisNode (int) to-node thatNode (int).
-
Patterns
-
int add-pattern with-inputs theseInputs (list) with-outputs theseOutputs (list).
-
int delete-pattern at-position thisPosition (int).
-
list get-all-patterns.
-
initializing
-
int initialize-network with-init-function
thisFunction = "Randomize_Weights" (string) with-init-parameters
theseParameters = {-1,1} (list).
-
training
-
list train with-function thisFunction =
"Std_BackPropagation" (string) for-epochs thisManytimes = 1 (int)
with-parameters theseParameters = {.2} (list).
-
float test.
-
running
-
list run with-inputs theseInputs (list).
-
files
-
int load-patterns from-file "thisFile".
-
int save-patterns to-file "thisFile".
-
int load-network from-file "thisFile".
-
int save-network to-file "thisFile".