- Add similarity measuring methods

AtlasMl/atlasml/ml/SimilarityMeasurement/Cosine.py

@@ -0,0 +1,35 @@
+import numpy as np
+from scipy.spatial.distance import cosine
+from AtlasMl.atlasml.ml.VectorEmbeddings.ModelDimension import ModelDimension
+
+
+def compute_cosine_similarity(embedding_vector, model: ModelDimension, comparison_vector):
+    """
+    Computes the cosine similarity between two embedding vectors.
+
+    Parameters:
+        embedding_vector (iterable): The input vector.
+        model (ModelDimension): An enum member indicating the dimension of the embedding_vector.
+        comparison_vector (iterable): The second embedding vector to compare with.
+
+    Returns:
+        float: The cosine similarity between the two vectors.
+    """
+    # Convert inputs to numpy arrays (in case they aren't already)
+    emb1 = np.array(embedding_vector)
+    emb2 = np.array(comparison_vector)
+
+    # Ensure both vectors have the same shape
+    if emb1.shape != emb2.shape:
+        raise ValueError("Both vectors must have the same dimensions.")
+    # TODO: Add vector re-shaper
+
+    # Calculate cosine similarity
+    similarity = 1.0 - cosine(emb1, emb2)
+    return similarity
+
+
+# Sanity Check
+example_vec1 = [1, 2, 3]
+example_vec2 = [4, 5, 6]
+computed_distance = compute_cosine_similarity(example_vec1, ModelDimension.text_embedding_three_small, example_vec2)

AtlasMl/atlasml/ml/SimilarityMeasurement/Euclidian.py

@@ -0,0 +1,35 @@
+import numpy as np
+from scipy.spatial.distance import euclidean
+from AtlasMl.atlasml.ml.VectorEmbeddings.ModelDimension import ModelDimension
+
+
+def compute_euclidean_distance(embedding_vector, model: ModelDimension, comparison_vector):
+    """
+    Computes the Euclidean distance between two embedding vectors.
+
+    Parameters:
+        embedding_vector (iterable): The input vector.
+        model (ModelDimension): An enum member indicating the dimension of the embedding_vector.
+        comparison_vector (iterable): The second embedding vector to compare with.
+
+    Returns:
+        float: The Euclidean distance between the two vectors.
+    """
+    # Convert inputs to numpy arrays (in case they aren't already)
+    vec1 = np.array(embedding_vector)
+    vec2 = np.array(comparison_vector)
+
+    # Ensure both vectors have the same shape
+    if vec1.shape != vec2.shape:
+        raise ValueError("Both vectors must have the same dimensions.")
+    # TODO: Add vector re-shaper
+
+    # Calculate Euclidean distance
+    distance = euclidean(vec1, vec2)
+    return distance
+
+
+# Sanity Check
+example_vec1 = [1, 2, 3]
+example_vec2 = [4, 5, 6]
+computed_distance = compute_euclidean_distance(example_vec1, ModelDimension.text_embedding_three_small, example_vec2)

AtlasMl/atlasml/ml/SimilarityMeasurement/Jaccard.py

@@ -0,0 +1,36 @@
+import numpy as np
+from scipy.spatial.distance import jaccard
+from AtlasMl.atlasml.ml.VectorEmbeddings.ModelDimension import ModelDimension
+
+
+def compute_jaccard_similarity(embedding_vector, model: ModelDimension, comparison_vector):
+    """
+    Computes the Jaccard similarity between two vectors.
+
+    Parameters:
+        embedding_vector (iterable): The input vector.
+        model (ModelDimension): An enum member indicating the dimension of the embedding_vector.
+        comparison_vector (iterable): The second embedding vector to compare with.
+
+    Returns:
+        float: The Jaccard similarity between the two vectors.
+    """
+    # Convert inputs to numpy arrays
+    vec1 = np.array(embedding_vector)
+    vec2 = np.array(comparison_vector)
+
+    # Ensure both vectors have the same shape
+    if vec1.shape != vec2.shape:
+        raise ValueError("Both vectors must have the same dimensions.")
+    # TODO: Add vector re-shaper
+
+    # Calculate Jaccard distance then convert to similarity
+    distance = jaccard(vec1, vec2)
+    similarity = 1.0 - distance
+    return similarity
+
+
+# Sanity Check
+example_vec1 = [1, 2, 3]
+example_vec2 = [4, 5, 6]
+computed_distance = compute_jaccard_similarity(example_vec1, ModelDimension.text_embedding_three_small, example_vec2)

AtlasMl/atlasml/ml/VectorEmbeddings/ModelDimension.py

@@ -0,0 +1,6 @@
+from enum import Enum
+
+class ModelDimension(Enum):
+    text_embedding_three_small = 1536
+    text_embedding_three_large = 3072
+    # Add more models that are to be used and their vector dimensions