Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
nokuno
committed
Sep 1, 2011
1 parent
a9bc9a5
commit 33f8cfd
Showing
5 changed files
with
207 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,11 @@ | ||
#!/bin/bash | ||
if [ $# != '2' ] | ||
then | ||
echo "local.sh input_path output_path" | ||
exit | ||
fi | ||
|
||
for i in `seq 1 10`; do | ||
cat $1 | ./perceptron.py -w $2/weight`expr $i - 1`.txt | sort | ./reduce.py > $2/weight$i.txt; | ||
done; | ||
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,89 @@ | ||
#!/usr/bin/env python | ||
#encoding: utf-8 | ||
|
||
from sys import stdin | ||
from optparse import OptionParser | ||
from collections import defaultdict | ||
from random import shuffle, seed | ||
|
||
|
||
def parse(line): | ||
pair = line.strip().split(" ", 1) | ||
if len(pair) != 2: return None | ||
label, document = pair | ||
|
||
features = defaultdict(float) | ||
for feature in document.split(" "): | ||
key, value = feature.split(":", 1) | ||
features[key] = float(value) | ||
|
||
return (label, features) | ||
|
||
def dot(x, y): | ||
return sum(x.get(i, 0.) * y[i] for i in y.iterkeys()) | ||
|
||
class Perceptron: | ||
def __init__(self, weight): | ||
self.weight = weight | ||
|
||
def train(self, documents, iteration, learn): | ||
for label, features in documents: | ||
if not label in self.weight: | ||
self.weight[label] = defaultdict(float) | ||
for i in range(iteration): | ||
shuffle(documents) | ||
for label, features in documents: | ||
# prediction | ||
prediction, scores = self.predict(features) | ||
|
||
# update | ||
if prediction != label: | ||
for key, value in features.iteritems(): | ||
self.weight[label][key] += learn * value | ||
self.weight[prediction][key] -= learn * value | ||
|
||
|
||
def predict(self, features): | ||
max_label = None | ||
scores = {} | ||
for label, weight in self.weight.iteritems(): | ||
scores[label] = dot(weight, features) | ||
if max_label == None or scores[label] > scores[max_label]: | ||
max_label = label | ||
return (max_label, scores) | ||
|
||
def output(self): | ||
result = "" | ||
for label, weight in self.weight.iteritems(): | ||
weight_str = " ".join(key + ":" + str(value) for key, value in weight.iteritems()) | ||
result += label + " " + weight_str + "\n" | ||
return result | ||
|
||
if __name__ == '__main__': | ||
parser = OptionParser() | ||
parser.add_option("-i", dest="iteration", type="int", default=1) | ||
parser.add_option("-e", dest="eta", type="float", default=0.001) | ||
parser.add_option("-w", dest="weight") | ||
(options, args) = parser.parse_args() | ||
|
||
# Load file | ||
documents = [] | ||
for line in stdin: | ||
documents.append(parse(line)) | ||
|
||
# Load weights from Distributed Cache | ||
weight = defaultdict(lambda: defaultdict(float)) | ||
if options.weight != None: | ||
for line in open(options.weight): | ||
parsed = parse(line) | ||
if parsed == None: continue | ||
key, value = parsed | ||
weight[key] = value | ||
|
||
# Train perceptron | ||
perceptron = Perceptron(weight) | ||
perceptron.train(documents, options.iteration, options.eta) | ||
|
||
# Output weight file | ||
print perceptron.output(), | ||
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,36 @@ | ||
#!/usr/bin/env python | ||
#encoding: utf-8 | ||
|
||
from perceptron import * | ||
|
||
if __name__ == '__main__': | ||
parser = OptionParser() | ||
parser.add_option("-w", dest="weight") | ||
(options, args) = parser.parse_args() | ||
|
||
# Load file | ||
documents = [] | ||
for line in stdin: | ||
documents.append(parse(line)) | ||
|
||
# Load weights from Distributed Cache | ||
weight = defaultdict(lambda: defaultdict(float)) | ||
if options.weight != None: | ||
for line in open(options.weight): | ||
parsed = parse(line) | ||
if parsed == None: continue | ||
key, value = parsed | ||
weight[key] = value | ||
|
||
perceptron = Perceptron(weight) | ||
|
||
# Test prediction | ||
correct = 0. | ||
for label, features in documents: | ||
prediction, scores = perceptron.predict(features) | ||
if prediction == label: | ||
correct += 1. | ||
|
||
# Output accuracy | ||
print correct / len(documents) | ||
|
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,41 @@ | ||
#!/usr/bin/env python | ||
from sys import stdin | ||
from optparse import OptionParser | ||
from collections import defaultdict | ||
|
||
def reduce(label, values): | ||
weight = defaultdict(float) | ||
|
||
# summation | ||
for value in values: | ||
for feature in value.split(" "): | ||
pair = feature.split(":") | ||
if len(pair) != 2: continue | ||
k, v = pair | ||
weight[k] += float(v) | ||
|
||
# averaging | ||
for k in weight.iterkeys(): | ||
weight[k] /= len(values) | ||
|
||
weight_str = " ".join(k + ":" + str(v) for k, v in weight.iteritems()) | ||
print "%s %s" % (label, weight_str) | ||
|
||
if __name__ == "__main__": | ||
current = None | ||
values = [] | ||
|
||
for line in stdin: | ||
splited = line.strip().split(" ", 1) | ||
if len(splited) != 2: | ||
continue | ||
key, value = splited | ||
|
||
if current != None and current != key: | ||
reduce(current, values) | ||
values = [] | ||
|
||
current = key | ||
values += [value] | ||
|
||
reduce(current, values) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,30 @@ | ||
#!/bin/bash | ||
if [ $# != '2' ] | ||
then | ||
echo "stream.sh input_path output_path" | ||
exit | ||
fi | ||
|
||
# First iteration | ||
echo "Iteration 1" | ||
hadoop jar \ | ||
$HADOOP_HOME/hadoop-streaming.jar \ | ||
-files perceptron.py,reduce.py \ | ||
-input $1 \ | ||
-output $2/weight1 \ | ||
-mapper ./perceptron.py \ | ||
-reducer ./reduce.py | ||
|
||
# Run iteration | ||
for i in `seq 2 10` | ||
do | ||
echo "Iteration $i" | ||
hadoop jar \ | ||
$HADOOP_HOME/hadoop-streaming.jar \ | ||
-files perceptron.py,reduce.py,hdfs:///user/yookuno/$2/weight`expr $i - 1`/part-00000 \ | ||
-input $1 \ | ||
-output $2/weight$i \ | ||
-mapper "./perceptron.py -w part-00000" \ | ||
-reducer ./reduce.py | ||
done | ||
|