#include <lwpr.h>
Data Fields | |
| int | nIn |
| Number N of input dimensions. | |
| int | nInStore |
| Storage-size of any N-vector, for aligment purposes. | |
| int | nOut |
| Number M of output dimensions. | |
| int | n_data |
| Number of training data the model has seen. | |
| double * | mean_x |
| Mean of all training data the model has seen (Nx1). | |
| double * | var_x |
| Mean of all training data the model has seen (Nx1). | |
| char * | name |
| An optional description of the model (Mx1). | |
| int | diag_only |
| Flag that determines whether distance matrices are handled as diagonal-only. | |
| int | meta |
| Flag that determines wheter 2nd order updates to LWPR_ReceptiveField.M are computed. | |
| double | meta_rate |
| Learning rate for 2nd order updates. | |
| double | penalty |
| Penalty factor used within distance metric updates. | |
| double * | init_alpha |
| Initial learning rate for 2nd order distance metric updates (NxN). | |
| double * | norm_in |
| Input normalisation (Nx1). Adjust this to the expected variation of your data. | |
| double * | norm_out |
| Output normalisation. Adjust this to the expected variation of your output data. | |
| double * | init_D |
| Initial distance metric (NxN). This often requires some tuning (NxN). | |
| double * | init_M |
| Cholesky factorisation of LWPR_Model.init_D (NxN). | |
| double | w_gen |
| Threshold that determines the minimum activation before a new RF is created. | |
| double | w_prune |
| Threshold that determines above which (second highest) activation a RF is pruned. | |
| double | init_lambda |
| Initial forgetting factor. | |
| double | final_lambda |
| Final forgetting factor. | |
| double | tau_lambda |
| This parameter describes the annealing schedule of the forgetting factor. | |
| double | init_S2 |
| Initial value for sufficient statistics LWPR_ReceptiveField.SSs2. | |
| double | add_threshold |
| Threshold that determines when a new PLS regression axis is added. | |
| LWPR_Kernel | kernel |
| Describes which kernel function is used (Gaussian or BiSquare). | |
| int | update_D |
| Flag that determines whether distance metric updates are performed (default: 1). | |
| LWPR_SubModel * | sub |
| Array of SubModels, one for each output dimension. | |
| struct LWPR_Workspace * | ws |
| Array of Workspaces, one for each thread (cf. LWPR_NUM_THREADS). | |
| double * | storage |
| Pointer to allocated memory. Do not touch. | |
| double * | xn |
| Used to hold a normalised input vector (Nx1). | |
| double * | yn |
| Used to hold a normalised output vector (Nx1). | |
| int | isPersistent |
| MEX-specific flag which determines whether this LWPR_Model is persistent. | |
This structure contains flags and initial values that determine the behaviour of the LWPR algorithm, and also provides some statistics about the model.
It should always be initialised with lwpr_init_model, and destroyed with lwpr_free_model. Note that both functions do not allocate/free the space for the LWPR_Model itself.
MEX-specific flag which determines whether this LWPR_Model is persistent.
This variable is only included in the LWPR_Model structure if the library is compiled with the directive MATLAB (i.e. for MEX-file usage). In that case, isPersistent=1 indicates that the LWPR model should be protected from automatic memory cleanups as performed by MATLAB.
1.5.9