diff --git a/README.md b/README.md
index 35eebe9..b6afd92 100644
--- a/README.md
+++ b/README.md
@@ -6,4 +6,4 @@ Fast, GPU-friendly, differentiable computation of Intrinsic Dimension via Maximu
 
 ### References
 - [E. Levina, P. Bickel; "Maximum Likelihood Estimation of Intrinsic Dimension", Advances in Neural Information Processing Systems, 2004](https://papers.nips.cc/paper_files/paper/2004/hash/74934548253bcab8490ebd74afed7031-Abstract.html)
-- [E. Facco, M. d'Errico, A. Rodriguez, A. Laio; "Estimating the intrinsic dimension of datasets by a minimal neighborhood information", Mature Scientific Reports, 2017](https://www.nature.com/articles/s41598-017-11873-y)
+- [E. Facco, M. d'Errico, A. Rodriguez, A. Laio; "Estimating the intrinsic dimension of datasets by a minimal neighborhood information", Nature Scientific Reports, 2017](https://www.nature.com/articles/s41598-017-11873-y)
diff --git a/fastwonn/mle.py b/fastwonn/mle.py
index 4df825f..719bc2e 100755
--- a/fastwonn/mle.py
+++ b/fastwonn/mle.py
@@ -1,69 +1,71 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-# ──────────────────────────────────────────────────────────────────────────────
-from contextlib import ExitStack
-
-import torch
-from torch import Tensor
-
-from .impl import call_to_impl_cdist_topk
-
-# ──────────────────────────────────────────────────────────────────────────────
-__all__ = ["mle_id", "mle_id_avg"]
-
-
-# ──────────────────────────────────────────────────────────────────────────────
-def mle_id(
-    x: Tensor,
-    nneigh: int = 2,
-    twonn_fix: bool = False,
-    differentiable: bool = False,
-    impl: str = "torch",
-) -> Tensor:
-
-    with ExitStack() as stack:
-        stack.enter_context(torch.no_grad()) if not differentiable else None
-
-        ks: Tensor = call_to_impl_cdist_topk[impl](x, nneigh, False)[:, 1:]
-
-        if twonn_fix and nneigh == 2:
-            return -2 * ks.size(0) / torch.log(torch.div(*torch.unbind(ks, 1))).sum()
-
-        return (2 * (nneigh - 1) / torch.log(ks[:, -1].view(-1, 1) / ks).sum(1)).mean()
-
-
-# ──────────────────────────────────────────────────────────────────────────────
-
-
-def mle_id_avg(
-    x: Tensor,
-    nneigh_min: int = 2,
-    nneigh_max: int = 10,
-    twonn_fix: bool = False,
-    differentiable: bool = False,
-    impl: str = "torch",
-) -> Tensor:
-
-    with ExitStack() as stack:
-        stack.enter_context(torch.no_grad()) if not differentiable else None
-
-        ks: Tensor = call_to_impl_cdist_topk[impl](x, nneigh_max, False)[:, 1:]
-        runs = [
-            (
-                2
-                * (nneigh_max - 1 - i)
-                / torch.log(
-                    ks[:, -1 - i].view(-1, 1) / (ks[:, :-i] if i != 0 else ks)
-                ).sum(1)
-            ).mean()
-            for i in range(nneigh_max - nneigh_min + (not twonn_fix))
-        ]
-
-        if twonn_fix and nneigh_min == 2:
-            runs.append(
-                -2
-                * ks.size(0)
-                / torch.log(torch.div(*torch.unbind(ks[:, 0:2], 1))).sum()
-            )
-
-        return torch.stack(runs).nanmean()
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+# ──────────────────────────────────────────────────────────────────────────────
+from contextlib import ExitStack
+
+import torch
+from torch import Tensor
+
+from .impl import call_to_impl_cdist_topk
+
+# ──────────────────────────────────────────────────────────────────────────────
+__all__ = ["mle_id", "mle_id_avg"]
+
+
+# ──────────────────────────────────────────────────────────────────────────────
+def mle_id(
+    x: Tensor,
+    nneigh: int = 2,
+    twonn_fix: bool = False,
+    differentiable: bool = False,
+    impl: str = "torch",
+) -> Tensor:
+
+    with ExitStack() as stack:
+        stack.enter_context(torch.no_grad()) if not differentiable else None
+
+        ks: Tensor = call_to_impl_cdist_topk[impl](x, nneigh, False)[:, 1:]
+
+        if twonn_fix and nneigh == 2:
+            return -2 * ks.size(0) / torch.log(torch.div(*torch.unbind(ks, 1))).sum()
+
+        return (2 * (nneigh - 1) / torch.log(ks[:, -1].view(-1, 1) / ks).sum(1)).mean()
+
+
+# ──────────────────────────────────────────────────────────────────────────────
+
+
+def mle_id_avg(
+    x: Tensor,
+    nneigh_min: int = 2,
+    nneigh_max: int = 10,
+    twonn_fix: bool = False,
+    differentiable: bool = False,
+    impl: str = "torch",
+) -> Tensor:
+
+    with ExitStack() as stack:
+        stack.enter_context(torch.no_grad()) if not differentiable else None
+
+        twonn_sep: bool = twonn_fix and nneigh_min == 2
+
+        ks: Tensor = call_to_impl_cdist_topk[impl](x, nneigh_max, False)[:, 1:]
+        runs = [
+            (
+                2
+                * (nneigh_max - 1 - i)
+                / torch.log(
+                    ks[:, -1 - i].view(-1, 1) / (ks[:, :-i] if i != 0 else ks)
+                ).sum(1)
+            ).mean()
+            for i in range(nneigh_max - nneigh_min + (not twonn_sep))
+        ]
+
+        if twonn_sep:
+            runs.append(
+                -2
+                * ks.size(0)
+                / torch.log(torch.div(*torch.unbind(ks[:, 0:2], 1))).sum()
+            )
+
+        return torch.stack(runs).nanmean()
diff --git a/setup.py b/setup.py
index dd2f71a..8d5de1d 100755
--- a/setup.py
+++ b/setup.py
@@ -22,7 +22,7 @@ def read(fname):
 
 setup(
     name=PACKAGENAME,
-    version="0.0.7",
+    version="0.0.8",
     author="Emanuele Ballarin",
     author_email="emanuele@ballarin.cc",
     url="https://github.com/emaballarin/fastwonn",