my-asu-archive/CSE515_MWDB_Project
1
0
Fork 0
mirror of https://github.com/20kaushik02/CSE515_MWDB_Project.git synced 2025-12-06 11:14:06 +00:00
Code Issues Packages Projects Releases Wiki Activity

nmf update, use own svd nmf

Browse Source
This commit is contained in:
Kaushik Narayan R 2023-10-14 12:30:18 -07:00
parent 431f86b2a5
commit 7c0c4a1838
2 changed files with 24 additions and 39 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@ -3,3 +3,4 @@ Other code/
*.zip *.zip
*.env *.env
__pycache__ __pycache__
*.json

48
Phase 2/utils.py
View File

@ -5,10 +5,9 @@ import random
import cv2 import cv2
import numpy as np import numpy as np
from scipy.stats import pearsonr from scipy.stats import pearsonr
from scipy.sparse.linalg import svds # from scipy.sparse.linalg import svds
from sklearn.decomposition import NMF # from sklearn.decomposition import NMF
from sklearn.decomposition import LatentDirichletAllocation from sklearn.decomposition import LatentDirichletAllocation
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
# from sklearn.cluster import KMeans # from sklearn.cluster import KMeans
@ -817,13 +816,21 @@ def svd(matrix, k):
return left_singular_vectors, np.diag(singular_values), right_singular_vectors.T return left_singular_vectors, np.diag(singular_values), right_singular_vectors.T
def nmf(matrix, k, num_iterations=100):
def nmf(matrix, k, H=None, update_H=True, num_iterations=100):
"""
Non-negative matrix factorization by multiplicative update
Pass `H` and `update_H=False` to transform given data as per the given H matrix, else leave `H=None` and `update_H=True` to fit and transform
"""
d1, d2 = matrix.shape d1, d2 = matrix.shape
# Initialize W and H matrices with random non-negative values # Initialize W and H matrices with random non-negative values
W = np.random.rand(d1, k) W = np.random.rand(d1, k)
if update_H is True:
H = np.random.rand(k, d2) H = np.random.rand(k, d2)
for iteration in range(num_iterations): for iteration in range(num_iterations):
if update_H is True:
# Update H matrix # Update H matrix
numerator_h = np.dot(W.T, matrix) numerator_h = np.dot(W.T, matrix)
denominator_h = np.dot(np.dot(W.T, W), H) denominator_h = np.dot(np.dot(W.T, W), H)
@ -836,6 +843,7 @@ def nmf(matrix, k, num_iterations=100):
return W, H return W, H
def extract_latent_semantics_from_feature_model( def extract_latent_semantics_from_feature_model(
fd_collection, fd_collection,
k, k,
@ -879,9 +887,8 @@ def extract_latent_semantics_from_feature_model(
match valid_dim_reduction_methods[dim_reduction_method]: match valid_dim_reduction_methods[dim_reduction_method]:
# singular value decomposition # singular value decomposition
# sparse version of SVD to get only k singular values
case 1: case 1:
U, S, V_T = svds(feature_vectors, k=k) U, S, V_T = svd(feature_vectors, k=k)
all_latent_semantics = { all_latent_semantics = {
"image-semantic": U.tolist(), "image-semantic": U.tolist(),
@ -906,18 +913,7 @@ def extract_latent_semantics_from_feature_model(
min_value = np.min(feature_vectors) min_value = np.min(feature_vectors)
feature_vectors_shifted = feature_vectors - min_value feature_vectors_shifted = feature_vectors - min_value
model = NMF( W, H = nmf(feature_vectors_shifted, k)
n_components=k,
init="random",
solver="cd",
alpha_H=0.01,
alpha_W=0.01,
max_iter=10000,
)
model.fit(feature_vectors_shifted)
W = model.transform(feature_vectors_shifted)
H = model.components_
all_latent_semantics = { all_latent_semantics = {
"image-semantic": W.tolist(), "image-semantic": W.tolist(),
@ -1053,9 +1049,8 @@ def extract_latent_semantics_from_sim_matrix(
match valid_dim_reduction_methods[dim_reduction_method]: match valid_dim_reduction_methods[dim_reduction_method]:
# singular value decomposition # singular value decomposition
# sparse version of SVD to get only k singular values
case 1: case 1:
U, S, V_T = svds(feature_vectors, k=k) U, S, V_T = svd(feature_vectors, k=k)
all_latent_semantics = { all_latent_semantics = {
"image-semantic": U.tolist(), "image-semantic": U.tolist(),
@ -1080,18 +1075,7 @@ def extract_latent_semantics_from_sim_matrix(
min_value = np.min(feature_vectors) min_value = np.min(feature_vectors)
feature_vectors_shifted = feature_vectors - min_value feature_vectors_shifted = feature_vectors - min_value
model = NMF( W, H = nmf(feature_vectors_shifted, k)
n_components=k,
init="random",
solver="cd",
alpha_H=0.01,
alpha_W=0.01,
max_iter=10000,
)
model.fit(feature_vectors_shifted)
W = model.transform(feature_vectors_shifted)
H = model.components_
all_latent_semantics = { all_latent_semantics = {
"image-semantic": W.tolist(), "image-semantic": W.tolist(),