|
SPRAAK
|
All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Groups Pages
|
SPRAAK
|
a connection between two layers More...
Collaboration diagram for SprMlpConnect:Data Fields | |
| SprMlpLayer * | src |
| the 'from/source' layer More... | |
| SprMlpLayer * | dst |
| the 'to/destination' layer More... | |
| SprMlpArg * | args |
| extra arguments More... | |
| char * | ndx_fname |
| file with indices (reduced/tree) More... | |
| int | nndxs_x |
| remember if the indices are stored in a matrix More... | |
| int | nndxs_y |
| remember if the indices are stored in a matrix More... | |
| char * | param_fname |
| file with parameters More... | |
| int | nparams_x |
| remember if params are stored in a matrix More... | |
| int | nparams_y |
| remember if params are stored in a matrix More... | |
| int | nr_arcs |
| the number of arcs (with specific parameters) More... | |
| int | type |
| type of connection More... | |
| int | dtype_fwd |
| derived type of connection More... | |
| int | dtype_bwd |
| derived type of connection More... | |
| unsigned long long | tree_cnt |
| count the number of tree evaluations More... | |
| int | xfun_len1 |
| length of the function list - 1 More... | |
| int | node_len |
| nr. of intermediate nodes needed by the sequence of functions More... | |
| SprMlpFun * | fun_list |
| the list of (non-linear) operations on every arc More... | |
| int | size_in |
| size of the input layer (including bias) More... | |
| int | size_out |
| size of the output layer More... | |
| SprNNFloatData * | ptr_in |
| pointer to the nodes of the input layer (including the bias) More... | |
| SprNNFloatData * | ptr_out |
| pointer to the nodes of the output layer More... | |
| int | dE_offs |
| offset to the error gradient for the input nodes More... | |
| int | param_i |
| extra parameter (integer) More... | |
| SprNNFloatCalc | param_f |
| extra parameter (float) More... | |
| SprNNFloatData * | weight |
| connection with another layer (tree-connection) More... | |
| int | flags |
| extra info such as use of bias, tree evaluation mode, ... More... | |
| int | nparams |
| SprNNFloatData * | param |
| parameters [and incremental updates during training] More... | |
| int * | ndx |
| all indices (reduced/tree) More... | |
| SprNNFloatData * | node |
| nodes for temporary results and for partial derivatives More... | |
| int * | offset |
| SprMlpXFun * | fun_fwd |
| array of functions on the arcs for forward evaluation More... | |
| union _Union4_MLP_MAIN_ | fwd |
| forward evaluation of all relevant arcs More... | |
| SprMlpXFun * | fun_bwd |
| array of functions on the arcs for backward evaluation More... | |
| union _Union5_MLP_MAIN_ | bwd |
| backward training of all relevant arcs More... | |
| struct t_mlp_connect * | next |
| next set of connections More... | |
| struct t_mlp_connect * | prev |
| previous set of connections More... | |
| struct t_mlp_connect * | tree_fwd |
| tree-structured evaluation (forward evaluation) More... | |
| struct t_mlp_connect * | tree_next |
| struct t_mlp_connect * | tree_bwd |
| tree-structured evaluation (backward evaluation) More... | |
| struct t_mlp_connect * | tree_prev |
a connection between two layers
| SprMlpLayer* SprMlpConnect::src |
the 'from/source' layer
| SprMlpLayer* SprMlpConnect::dst |
the 'to/destination' layer
| SprMlpArg* SprMlpConnect::args |
extra arguments
| char* SprMlpConnect::ndx_fname |
file with indices (reduced/tree)
| int SprMlpConnect::nndxs_x |
remember if the indices are stored in a matrix
| int SprMlpConnect::nndxs_y |
remember if the indices are stored in a matrix
| char* SprMlpConnect::param_fname |
file with parameters
| int SprMlpConnect::nparams_x |
remember if params are stored in a matrix
| int SprMlpConnect::nparams_y |
remember if params are stored in a matrix
| int SprMlpConnect::nr_arcs |
the number of arcs (with specific parameters)
| int SprMlpConnect::type |
type of connection
| int SprMlpConnect::dtype_fwd |
derived type of connection
| int SprMlpConnect::dtype_bwd |
derived type of connection
| unsigned long long SprMlpConnect::tree_cnt |
count the number of tree evaluations
| int SprMlpConnect::xfun_len1 |
length of the function list - 1
| int SprMlpConnect::node_len |
nr. of intermediate nodes needed by the sequence of functions
| SprMlpFun* SprMlpConnect::fun_list |
the list of (non-linear) operations on every arc
| int SprMlpConnect::size_in |
size of the input layer (including bias)
| int SprMlpConnect::size_out |
size of the output layer
| SprNNFloatData* SprMlpConnect::ptr_in |
pointer to the nodes of the input layer (including the bias)
| SprNNFloatData* SprMlpConnect::ptr_out |
pointer to the nodes of the output layer
| int SprMlpConnect::dE_offs |
offset to the error gradient for the input nodes
| int SprMlpConnect::param_i |
extra parameter (integer)
| SprNNFloatCalc SprMlpConnect::param_f |
extra parameter (float)
| SprNNFloatData* SprMlpConnect::weight |
connection with another layer (tree-connection)
| int SprMlpConnect::flags |
extra info such as use of bias, tree evaluation mode, ...
| int SprMlpConnect::nparams |
total number of parameters -> needed to find deltas in the param array
| SprNNFloatData* SprMlpConnect::param |
parameters [and incremental updates during training]
| int* SprMlpConnect::ndx |
all indices (reduced/tree)
| SprNNFloatData* SprMlpConnect::node |
nodes for temporary results and for partial derivatives
| int* SprMlpConnect::offset |
allocated block, offset to param/ndx/node per input node (needed for tree structured evaluation)
| SprMlpXFun* SprMlpConnect::fun_fwd |
array of functions on the arcs for forward evaluation
| union _Union4_MLP_MAIN_ SprMlpConnect::fwd |
forward evaluation of all relevant arcs
| SprMlpXFun* SprMlpConnect::fun_bwd |
array of functions on the arcs for backward evaluation
| union _Union5_MLP_MAIN_ SprMlpConnect::bwd |
backward training of all relevant arcs
| struct t_mlp_connect* SprMlpConnect::next |
next set of connections
| struct t_mlp_connect* SprMlpConnect::prev |
previous set of connections
| struct t_mlp_connect* SprMlpConnect::tree_fwd |
tree-structured evaluation (forward evaluation)
| struct t_mlp_connect* SprMlpConnect::tree_next |
next set of connections for a tree-structured evaluation (forward evaluation)
| struct t_mlp_connect* SprMlpConnect::tree_bwd |
tree-structured evaluation (backward evaluation)
| struct t_mlp_connect* SprMlpConnect::tree_prev |
previous set of connections for a tree-structured evaluation (backward evaluation)
1.8.6
