Update README.md

Author: yurymalkov

README.md

@@ -1,26 +1,19 @@
 # HNSW - Approximate nearest neighbor search
-Paper code for the HNSW 200M SIFT experiment and a header-only C++ HNSW implementation with python bindings.
+Header-only C++ HNSW implementation with python bindings. Paper code for the HNSW 200M SIFT experiment
 
-NEW: Added simple python bindings with incremental construction
+NEW: Added support for cosine similarity and inner product distances
 
 
-#### Test reproduction steps:
-To download and extract the bigann dataset:
-```bash
-python3 download_bigann.py
-```
-To compile:
-```bash
-cmake .
-make all
-```
+Part of the nmslib project https://github.com/nmslib/nmslib
 
-To run the test on 200M SIFT subset:
-```bash
-./main
-```
 
-The size of the bigann subset (in millions) is controlled by the variable **subset_size_milllions** hardcoded in **sift_1b.cpp**.
+
+Supported distances:
+1) Squared L2 ('l2')
+2) Inner product ('ip',the distance is 1.0 - $inner product$)
+3) Cosine similarity ('cosine', the same as the inner product, but vectors are normalized)
+
+For other spaces use the main library https://github.com/nmslib/nmslib 
 
 
 #### Python bindings example
@@ -36,7 +29,7 @@ data = np.float32(np.random.random((num_elements, dim)))
 data_labels = np.arange(num_elements)
 
 # Declaring index
-p = hnswlib.Index(space = 'l2', dim = dim) # Only l2 is supported currently
+p = hnswlib.Index(space = 'l2', dim = dim) # possible options are l2, cosine or ip
 
 # Initing index - the maximum number of elements should be known beforehand
 p.init_index(max_elements = num_elements, ef_construction = 200, M = 16)
@@ -45,7 +38,7 @@ p.init_index(max_elements = num_elements, ef_construction = 200, M = 16)
 int_labels = p.add_items(data, data_labels)
 
 # Controlling the recall by setting ef:
-p.set_ef(50)
+p.set_ef(50) # ef should always be > k
 
 # Query dataset, k - number of closest elements (returns 2 numpy arrays)
 labels, distances = p.knn_query(data, k = 1)
@@ -57,7 +50,25 @@ cd python_bindings
 python3 setup.py install
 ```
 
-The repo contrains parts of the Non-Metric Space Library's code https://github.com/searchivarius/nmslib
+#### 200M SIFT test reproduction steps:
+To download and extract the bigann dataset:
+```bash
+python3 download_bigann.py
+```
+To compile:
+```bash
+cmake .
+make all
+```
+
+To run the test on 200M SIFT subset:
+```bash
+./main
+```
+
+The size of the bigann subset (in millions) is controlled by the variable **subset_size_milllions** hardcoded in **sift_1b.cpp**.
+
+
 
 References:
 Malkov, Yu A., and D. A. Yashunin. "Efficient and robust approximate nearest neighbor search using Hierarchical Navigable Small World graphs." arXiv preprint arXiv:1603.09320 (2016).