add ultranest import

anesthetic/read/chain.py

@@ -3,6 +3,7 @@
 from anesthetic.read.getdist import read_getdist
 from anesthetic.read.cobaya import read_cobaya
 from anesthetic.read.multinest import read_multinest
+from anesthetic.read.ultranest import read_ultranest
 
 
 def read_chains(root, *args, **kwargs):
@@ -45,7 +46,11 @@ def read_chains(root, *args, **kwargs):
             "anesthetic.html#anesthetic.samples.MCMCSamples.remove_burn_in"
         )
     errors = []
-    for read in [read_polychord, read_multinest, read_cobaya, read_getdist]:
+    readers = [
+        read_polychord, read_multinest, read_cobaya,
+        read_ultranest, read_getdist
+    ]
+    for read in readers:
         try:
             samples = read(root, *args, **kwargs)
             samples.root = root

anesthetic/read/ultranest.py

@@ -0,0 +1,30 @@
+"""Read NestedSamples from ultranest results."""
+import os
+from anesthetic.samples import NestedSamples
+import h5py
+import json
+
+
+def read_ultranest(root, *args, **kwargs):
+    """Read <root>ev.dat and <root>phys_live.points in MultiNest format."""
+    with open(os.path.join(root, 'info', 'results.json')) as infofile:
+        labels = json.load(infofile)['paramnames']
+    num_params = len(labels)
+
+    filepath = os.path.join(root, 'results', 'points.hdf5')
+    with h5py.File(filepath, 'r') as fileobj:
+        points = fileobj['points']
+        _, ncols = points.shape
+        x_dim = ncols - 3 - num_params
+        logL_birth = points[:, 0]
+        logL = points[:, 1]
+        # u_samples = points[:,3:x_dim]
+        samples = points[:, 3 + x_dim:3 + x_dim + num_params]
+
+    kwargs['label'] = kwargs.get('label', os.path.basename(root))
+    columns = kwargs.pop('columns', labels)
+    labels = kwargs.pop('labels', labels)
+    data = samples
+
+    return NestedSamples(data=data, logL=logL, logL_birth=logL_birth,
+                         columns=columns, labels=labels, *args, **kwargs)

tests/example_data/makeun.py

@@ -0,0 +1,31 @@
+import numpy as np
+from ultranest import ReactiveNestedSampler
+
+ndim = 4
+sigma = np.logspace(-1, -6, ndim)
+width = 1 - 5 * sigma
+width[width < 1e-20] = 1e-20
+centers = (np.sin(np.arange(ndim) / 2.) * width + 1.) / 2.
+
+
+def loglike(theta):
+    """compute log-likelihood."""
+    like = - 0.5 * (((theta - centers) / sigma)**2).sum(axis=1) \
+           - 0.5 * np.log(2 * np.pi * sigma**2).sum()
+    return like
+
+
+def transform(x):
+    """transform according to prior."""
+    return x * 20 - 10
+
+
+paramnames = ['a', 'b', 'c', 'd']
+
+sampler = ReactiveNestedSampler(
+    paramnames, loglike, transform=transform,
+    log_dir='un', resume=True, vectorized=True)
+
+sampler.run(frac_remain=0.5, min_num_live_points=400)
+sampler.print_results()
+sampler.plot()

tests/example_data/un/info/post_summary.csv

@@ -0,0 +1,2 @@
+"a_mean","a_stdev","a_median","a_errlo","a_errup","b_mean","b_stdev","b_median","b_errlo","b_errup","c_mean","c_stdev","c_median","c_errlo","c_errup","d_mean","d_stdev","d_median","d_errlo","d_errup"
+5.025783329483370920e-01,9.929557150158441203e-02,5.066546971904539731e-01,4.016698354319299114e-01,6.012826682750720408e-01,7.370934893843076319e-01,2.131337783168813296e-03,7.370548529653415670e-01,7.350305173203235398e-01,7.393071000969335671e-01,9.206375311748793422e-01,4.603300006747625519e-05,9.206365431021019674e-01,9.205913816528905613e-01,9.206855255078583156e-01,9.987450408249887168e-01,1.018245828856401194e-06,9.987450545520015766e-01,9.987439938546689433e-01,9.987460211457328541e-01

tests/example_data/un/info/results.json

@@ -0,0 +1,77 @@
+{
+    "niter": 16650,
+    "logz": -12.414914594805953,
+    "logzerr": 0.7649516021328628,
+    "logz_bs": -12.378097297348035,
+    "logz_single": -12.414914594805953,
+    "logzerr_tail": 0.405716490463611,
+    "logzerr_bs": 0.6484944741256662,
+    "ess": 988.8766189106832,
+    "H": 38.639675527033575,
+    "Herr": 0.22373181701346842,
+    "posterior": {
+        "mean": [
+            0.5025783329483371,
+            0.7370934893843076,
+            0.9206375311748793,
+            0.9987450408249887
+        ],
+        "stdev": [
+            0.09929557150158441,
+            0.0021313377831688133,
+            4.6033000067476255e-05,
+            1.0182458288564012e-06
+        ],
+        "median": [
+            0.506654697190454,
+            0.7370548529653416,
+            0.920636543102102,
+            0.9987450545520016
+        ],
+        "errlo": [
+            0.4016698354319299,
+            0.7350305173203235,
+            0.9205913816528906,
+            0.9987439938546689
+        ],
+        "errup": [
+            0.601282668275072,
+            0.7393071000969336,
+            0.9206855255078583,
+            0.9987460211457329
+        ],
+        "information_gain_bits": [
+            3.7847253159407837,
+            4.14647919764959,
+            4.14647919764959,
+            4.14647919764959
+        ]
+    },
+    "maximum_likelihood": {
+        "logl": 28.526361154649237,
+        "point": [
+            0.4912994020358319,
+            0.7373000808358334,
+            0.9206344043582959,
+            0.9987452206617764
+        ],
+        "point_untransformed": [
+            0.5245649701017916,
+            0.5368650040417917,
+            0.5460317202179148,
+            0.5499372610330888
+        ]
+    },
+    "ncall": 39885,
+    "paramnames": [
+        "a",
+        "b",
+        "c",
+        "d"
+    ],
+    "logzerr_single": 0.31080410038734035,
+    "insertion_order_MWW_test": {
+        "independent_iterations": Infinity,
+        "converged": true
+    }
+}

tests/test_reader.py

@@ -11,6 +11,7 @@
 from anesthetic.read.getdist import read_getdist
 from anesthetic.read.cobaya import read_cobaya
 from anesthetic.read.multinest import read_multinest
+from anesthetic.read.ultranest import read_ultranest
 import pandas._testing as tm
 from anesthetic.read.hdf import HDFStore, read_hdf
 try:
@@ -184,6 +185,25 @@ def test_read_multinest():
     ns.plot_1d(['x0', 'x1', 'x2', 'x3'])
 
 
+def test_read_ultranest():
+    np.random.seed(3)
+    ns = read_ultranest('./tests/example_data/un')
+    params = ['a', 'b', 'c', 'd', 'logL', 'logL_birth', 'nlive']
+    assert_array_equal(ns.drop_labels().columns, params)
+    labels = ['a',
+              'b',
+              'c',
+              'd',
+              r'$\ln\mathcal{L}$',
+              r'$\ln\mathcal{L}_\mathrm{birth}$',
+              r'$n_\mathrm{live}$']
+    assert_array_equal(ns.get_labels(), labels)
+
+    assert isinstance(ns, NestedSamples)
+    ns.plot_2d(['a', 'b', 'c', 'd'])
+    ns.plot_1d(['a', 'b', 'c', 'd'])
+
+
 def test_read_polychord():
     np.random.seed(3)
     ns = read_polychord('./tests/example_data/pc')