ValueError: setting an array element with a sequence.Using pysparnn.matrix_distance.SlowEuclideanDistance

[liuchenbaidu]

import pysparnn.cluster_index as ci

from sklearn.feature_extraction.text import TfidfVectorizer
import pysparnn
data = [
'hello world',
'oh hello there',
'Play it',
'Play it again Sam',
]
data=['你在干什么',
'你在干啥子',
'你在做什么',
'你好啊',
'我喜欢吃香蕉']

tv = TfidfVectorizer()
tv.fit(data)

features_vec = tv.transform(data)
print(type(features_vec),features_vec.shape)

build the search index!

cp = ci.MultiClusterIndex(features_vec, data,pysparnn.matrix_distance.SlowEuclideanDistance)

search the index with a sparse matrix

search_data = [
'oh there',
'Play it again Frank'
]

search_data = [
'你在干啥','我喜欢吃香蕉'
]
search_features_vec = tv.transform(search_data)

res=cp.search(search_features_vec, k=3, k_clusters=3, return_distance=False)

print(res)

[kchaliki]

@liuchenbaidu indeed that code doesn't work with sparse matrices, the test actually uses dense which is why this went unnoticed. I did implement this separately somewhere using scikit's euclidean distance but it is so much slower than cosine that it begs the question whether you need it.