You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: src/components/Collections/SearchQuality/SearchQuality.jsx
+23-3Lines changed: 23 additions & 3 deletions
Original file line number
Diff line number
Diff line change
@@ -10,6 +10,24 @@ import { bigIntJSON } from '../../../common/bigIntJSON';
10
10
importEditorCommonfrom'../../EditorCommon';
11
11
import_from'lodash';
12
12
13
+
constexplainLog=`Example Output: [18/10/2024, 01:51:38] Point ID 1(1/100) precision@10: 0.8 (search time exact: 30ms, regular: 5ms)
14
+
15
+
Explanation:
16
+
This output compares the performance of an exact search (full kNN) versus an approximate search using ANN (Approximate Nearest Neighbor).
17
+
18
+
- precision@10: 0.8: Out of the top 10 results returned by the exact search, 8 were also found in the ANN search.
19
+
- Search Time: The exact search took 30ms, whereas the ANN search was much faster, taking only 5ms, but with a small loss in accuracy.
20
+
21
+
Tuning the HNSW Algorithm (Advanced mode):
22
+
- "hnsw_ef" parameter: Controls how many neighbors to consider during a search. Increasing "hnsw_ef" improves precision but slows down the search.
23
+
24
+
Practical Use:
25
+
- The ANN search (with HNSW) is significantly faster (5ms compared to 30ms) but slightly less accurate (precision@10: 0.8). You can adjust parameters like "hnsw_ef" in advanced mode to find the right balance between speed and accuracy.
26
+
27
+
Additional Tuning Parameters (need to be set during collection configuration):
28
+
1. "m" Parameter : Defines the number of edges per node in the graph. A higher "m" value improves accuracy but uses more memory.
29
+
2. "ef_construct" Parameter: Sets the number of neighbors to consider during index creation. A higher value increases precision but requires more time during indexing.`;
0 commit comments