Method work

hyperopt/tests/unit/test_tpe.py:471–508

(self, **kwargs)

        return self.LGMM1(samples, **self.kwargs)

    def work(self, **kwargs):
        self.__dict__.update(kwargs)
        self.worked = True
        samples = old_div(
            LGMM1(rng=self.rng, size=(self.n_samples,), **self.kwargs), self.q
        )
        # -- we've divided the LGMM1 by self.q to get ints here
        assert np.all(samples == samples.astype("int"))
        min_max = int(samples.min()), int(samples.max())
        print("SAMPLES RANGE", min_max)
        counts = np.bincount(samples.astype("int") - min_max[0])

        # print samples
        # print counts
        xcoords = np.arange(min_max[0], min_max[1] + 0.5) * self.q
        prob = np.exp(LGMM1_lpdf(xcoords, **self.kwargs))
        print(xcoords)
        print(prob)
        assert counts.sum() == self.n_samples
        y = old_div(counts, float(self.n_samples))

        if self.show:
            plt.scatter(xcoords, y, c="r", label="empirical")
            plt.scatter(xcoords, prob, c="b", label="predicted")
            plt.legend()
            plt.show()
        # -- calculate errors on the low end, don't take a mean
        #    over all the range spanned by a few outliers.
        err = ((prob - y) ** 2)[:20]
        print(np.max(err))
        print(np.mean(err))
        print(np.median(err))
        if self.show:
            raise nose.SkipTest()
        else:
            assert np.max(err) < 0.1
            assert np.mean(err) < 0.01
            assert np.median(err) < 0.01

    def test_basic_1(self):
        self.work(q=1)