Write to Existing
On the Output tab of the form, you have the option to write the results of the neural network to an existing file. This could be a block model that was previously created or it could be a set of points, such as sample locations. Estimates will be written to columns with the same names as the input field names. If the file already contains these fields the target fields will be cleared.
East, North, Z fields
Double-click (or click on the List icon) to select the coordinate fields in the Input file.
Additional Guide Fields
Additional guide fields can be used to help the training. These fields must be present in both the Input file and the Output file, which is why this option is not available when using the Create Output method. Only numeric data can be used as an Additional Guide Field.
The data in these Additional guide fields are used for training the Copilot in much the same way as the coordinate data. You can consider the guide fields as additional dimensions that the neural network will consider and learn the relationship between these data and the modelled attributes. It is recommended to only use data as guide fields if there is a decent relationship with the modelled attributes. You could use any kind of numerical data to help train if that data has information that is useful in predicting the attribute that you are modelling.
If, for example, you are modelling a shallow residual deposit such as a laterite or bauxite you may want to use Depth as an Additional Guide Field. This will mean that the neural network is trained on Eastings, Northings, RL, and the Depth data.
Block model file
If the selected output file is a block model file, with block dimension fields, discretisation can be used to produce a more reliable average. This is recommended, especially if the blocks are reasonably large compared to the size of the samples.
Specify the number of divisions in the East, North, and Z directions. These are integer values that control the number of discrete points to be queried in each direction within each block. The values of all the discretised points are then averaged to estimate a single value for the block, which is written to the output file.
For example, if you choose 2 East divisions, 2 North divisions, and 2 Z divisions, the application will obtain and average a total of 8 estimates for each block. This can obviously add significant time to the inference process if you have a large number of blocks. It is not advised to use small blocks and high discretisation values because querying so many points will be inefficient and may take a long time. The default is 1 so the only location that will be queried is at the centre of the block.
Overwrite target fields
Select this option to overwrite the contents of the Target fields when you run the function.
