'use strict';
import * as Utils from './Utils';
/**
* The inputs and targets that will be used to train and test network models.
*/
class DataSet {
/**
* Creates an instance of DataSet.
* @param {TFNestedArray} _inputs The cases that will be sent to TF's model
* fit(). The values must be nested arrays,
* because they will become TF tensors. See
* {@link https://js.tensorflow.org/api/latest/#tensor}
* for details. Note that the flat-array
* option is not yet supported.
* @param {ArrayOrder2} _targets The output values associated with the
* input data. Must be a two-dimensional
* nested array. Note: Proper format
* flexibility is coming soon!
* @example
* // create a data set with inputs and targets for three cases
* new tngs.DataSet([[ 0, 2, 0, 4 ], [ 9, 2, 9, 6 ], [ 3, 5, 7, 1 ]],
* [[ 1, 0, 0 ], [ 0, 1, 0 ], [ 0, 0, 1 ]]);
*
* // In this data, there are four inputs for each case. The targets are
* // one-hot classification probabilities, labeled "even", "mixed" and
* // "odd".
*/
constructor(_inputs, _targets) {
this._inputs = _inputs;
this._targets = _targets;
Utils.Assert(this._inputs.length > 0);
if (this._inputs.length !== this._targets.length) {
throw new Error('Data invalid. The number of inputs ('
+ this._inputs.length + ') does not match the '
+ 'number of targets (' + this._targets.length
+ ') .');
}
}
get inputs() { return this._inputs; }
get targets() { return this._targets; }
}
Object.freeze(DataSet);
export { DataSet };
//# sourceMappingURL=DataSet.js.map