gistfile1.txt

FROM ubuntu
MAINTAINER Hiroki Kumazaki

RUN echo "deb http://download.jubat.us/apt binary/" >> /etc/apt/sources.list
RUN apt-get update
RUN apt-get install -y --allow-unauthenticated  jubatus

EXPOSE 9199

ENV JUBATUS_HOME /opt/jubatus
ENV PATH /opt/jubatus/bin:${PATH}
ENV LD_LIBRARY_PATH /opt/jubatus/lib:${LD_LIBRARY_PATH}
ENV LDFLAGS -L$/opt/jubatus/lib ${LDFLAGS}
ENV CPLUS_INCLUDE_PATH /opt/jubatus/include:${CPLUS_INCLUDE_Path}

# set filename
ADD kmeans.json /tmp/config.json
CMD jubaclustering -f /tmp/config.json

jubatus-kmeans-sample.py

#!/usr/bin/env python
# coding: utf-8
import sys
import json
import random

import jubatus
from jubatus.common import Datum

host = '127.0.0.1'
port = 9199
name = 'test'

client = jubatus.Clustering(host, port, name)
clusters = [[], []]
for x in xrange(10000):
    if x & 1 == 0:
        # cluster 1
        d = Datum({
            'x' :100 + random.randint(-10, 10),
            'y' : 50 + random.randint(-20, 20)
        })
        clusters[0].append(d)
    else:
        # cluster 2
        d = Datum({
            'x' :-200 + random.randint(-10, 10),
            'y' : 10 + random.randint(-10, 10)
        })
        clusters[1].append(d)

for cluster in clusters:
    client.push(cluster)
    print("{size} data pushed".format(size=len(cluster)))

centers = client.get_k_center()

print("total {size} centers clustered".format(size=len(centers)))
for center in centers:
    print(center.num_values)

data = {'x': -130, 'y': -2}
nearest = client.get_nearest_center(Datum(data))
print("data nearest {d} is {n}".format(d=data, n=nearest.num_values))

kmeans.json

{
  "converter" : {
    "num_rules" : [
      { "key" : "*", "type" : "num" }
    ]
  },
  "parameter" : {
    "k" : 2,
    "compressor_method" : "compressive_kmeans",
    "bucket_size" : 1000,
    "compressed_bucket_size" : 100,
    "bicriteria_base_size" : 10,
    "bucket_length" : 4,
    "forgetting_factor" : 0,
    "forgetting_threshold" : 0.5
  },
  "method" : "kmeans"
}