GeorgeSeif

k = 10
n_init = 10
max_iter = 300
kmeans = faiss.Kmeans(d=data.shape[1], k=k, niter=max_iter, nredo=n_init, gpu=True)
kmeans.train(data.astype(np.float32))

e = time.time()
print("Training time = {}".format(e - s))

GeorgeSeif

clf = KMeans(n_clusters=10)

s = time.time()
clf.fit(data)
e = time.time()
print("Training time = {}".format(e - s))


s = time.time()
clf.predict(data)

GeorgeSeif

import numpy as np

data_size = 1000

data = np.random.normal((100, 100, 100), (20, 20, 20), (data_size, 3))

GeorgeSeif

s = time.time()

k = 10
n_init = 10
max_iter = 300
kmeans = faiss.Kmeans(d=x_train.shape[1], k=k, niter=max_iter, nredo=n_init)
kmeans.train(x_train.astype(np.float32))

e = time.time()
print("Training time = {}".format(e - s))

GeorgeSeif

clf = KMeans(n_clusters=10)

s = time.time()
clf.fit(x_train)
e = time.time()
print("Training time = {}".format(e - s))


s = time.time()
clf.predict(x_test)

GeorgeSeif

import numpy as np
import time

from sklearn.cluster import KMeans
from keras.datasets import mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()

x_train = x_train.reshape(len(x_train), -1).astype(float) / 255.
x_test = x_test.reshape(len(x_test), -1).astype(float) / 255.

GeorgeSeif

# CPU version only
conda install faiss-cpu -c pytorch

# Additional Python packages
pip3 install numpy
pip3 install scikit-learn
pip3 install tensorflow
pip3 install keras

GeorgeSeif

conda create --name faiss
conda activate faiss

GeorgeSeif

from sklearn.decomposition import LatentDirichletAllocation as LDA

NUM_TOPICS = 3

# Here we create and fit the LDA model
# The "document_word_matrix" is a 2D array where each row is a document
# and each column is a word. The cells contain the count of the word within
# each document
lda = LDA(n_components=NUM_TOPICS, n_jobs=-1)
lda.fit(document_word_matrix)

GeorgeSeif

import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer

def get_tf_idf(vectorizer):
  feature_names = vectorizer.get_feature_names()
  dense_vec = vectors.todense()
  dense_list = dense_vec.tolist()
  tfidf_data = pd.DataFrame(dense_list, columns=feature_names)
  return tfidf_data