[BUG] BORF failing without numba (#2254)

* fix #2245 * Empty commit for CI * try another solution to fix using njit and loop * fixes --------- Co-authored-by: MatthewMiddlehurst <[email protected]> Co-authored-by: Tony Bagnall <[email protected]> Co-authored-by: MatthewMiddlehurst <[email protected]>
aeon-toolkit · Oct 30, 2024 · b76073e · b76073e
1 parent 15d44e1
commit b76073e
Showing 1 changed file with 17 additions and 15 deletions.
diff --git a/aeon/transformations/collection/dictionary_based/_borf.py b/aeon/transformations/collection/dictionary_based/_borf.py
@@ -548,12 +548,15 @@ def _ndindex_2d_array(idx, dim2_shape):
 
 @nb.njit(cache=True)
 def _get_norm_bins(alphabet_size: int, mu=0, std=1):
-    return _ppf(np.linspace(0, 1, alphabet_size + 1)[1:-1], mu, std)
+    bins = []
+    for i in np.linspace(0, 1, alphabet_size + 1)[1:-1]:
+        bins.append(_ppf(i, mu, std))
+    return np.array(bins)
 
 
 @nb.njit(fastmath=True, cache=True)
 def _erfinv(x: float) -> float:
-    w = -math.log((1 - x) * (1 + x))
+    w = -np.log((1 - x) * (1 + x))
     if w < 5:
         w = w - 2.5
         p = 2.81022636e-08
@@ -566,7 +569,7 @@ def _erfinv(x: float) -> float:
         p = 0.246640727 + p * w
         p = 1.50140941 + p * w
     else:
-        w = math.sqrt(w) - 3
+        w = np.sqrt(w) - 3
         p = -0.000200214257
         p = 0.000100950558 + p * w
         p = 0.00134934322 + p * w
@@ -579,9 +582,9 @@ def _erfinv(x: float) -> float:
     return p * x
 
 
-@nb.vectorize(cache=True)
+@nb.njit(cache=True)
 def _ppf(x, mu=0, std=1):
-    return mu + math.sqrt(2) * _erfinv(2 * x - 1) * std
+    return mu + np.sqrt(2) * _erfinv(2 * x - 1) * std
 
 
 @nb.njit(fastmath=True, cache=True)
@@ -763,17 +766,16 @@ def _length(a):
 
 @nb.njit(cache=True)
 def _hash_function(v):
-
     byte_mask = np.uint64(255)
     bs = np.uint64(v)
     x1 = (bs) & byte_mask
-    x2 = (bs >> 8) & byte_mask
-    x3 = (bs >> 16) & byte_mask
-    x4 = (bs >> 24) & byte_mask
-    x5 = (bs >> 32) & byte_mask
-    x6 = (bs >> 40) & byte_mask
-    x7 = (bs >> 48) & byte_mask
-    x8 = (bs >> 56) & byte_mask
+    x2 = (bs >> np.uint64(8)) & byte_mask
+    x3 = (bs >> np.uint64(16)) & byte_mask
+    x4 = (bs >> np.uint64(24)) & byte_mask
+    x5 = (bs >> np.uint64(32)) & byte_mask
+    x6 = (bs >> np.uint64(40)) & byte_mask
+    x7 = (bs >> np.uint64(48)) & byte_mask
+    x8 = (bs >> np.uint64(56)) & byte_mask
 
     FNV_primer = np.uint64(1099511628211)
     FNV_bias = np.uint64(14695981039346656037)
@@ -800,7 +802,7 @@ def _hash_function(v):
 @nb.njit(cache=True)
 def _make_hash_table(ar):
     a = _length(len(ar))
-    mask = a - 1
+    mask = np.uint64(a - 1)
 
     uniques = np.empty(a, dtype=ar.dtype)
     uniques_cnt = np.zeros(a, dtype=np.int_)
@@ -822,7 +824,7 @@ def _set_item(uniques, uniques_cnt, mask, h, v, total, miss_hits, weight):
             break
         else:
             miss_hits += 1
-            index += 1
+            index += np.uint64(1)
             index = index & mask
     return total, miss_hits