上一期我们讲到Pycon 2016 tensorflow 研讨会总结 — tensorflow 手把手入门 #第一讲 . 今天是我们第二讲, 来趴一趴word2vec.
什么是word2vec?
用来学习文字向量表达的模型 (相关文本文字的的特征向量).
- 向量空间模型解决了NLP中数据稀疏问题, 如果文字是离散的. 即, 把文字映射到相邻的空间点上.
立刻上图感受一下word2vec:
这里看看与文字’Cat’接近的词汇, 一目了然啊~如果一定要给’cat’一个向量描述, 上图左边这一列特征和权重是不是挺合理的呢? 嘿嘿~~~
word2vec两种方法:
- 基于计数的(如, LSA)
- 预测型的: 试着用学习到的embeddings在相邻文字中预测文字(如, word2vec 和 其他神经概率语言模型)
Mikolov等人的NIPS论文, http://bit.ly/word2vec-paper
两种word2vec
- 连续Bag-of-Words (COBW)
- 从上下文来预测一个文字
- Skip-Gram
- 从一个文字来预测上下文
使得word2vec可扩展
- 使用对数回归把文字从假造的噪声文字中区分出来, 而不是使用完全的概率模型.
- 噪音对比估计(NCE) 损失.
- tf.nn.nce_loss()
- 用噪音文字扩展
Skip-Gram 模型(用目标文字预测上下文)
上下文/目标文字组合, 双向窗口大小为1:
the quick brown fox jumped over the lazy dog … →
([the, brown], quick), ([quick, fox], brown), ([brown,
jumped], fox),
输入/输出组合:
(quick, the), (quick, brown), (brown, quick), (brown,
fox), …
一般用SGD随机梯度下降优化
word2vec Tensorflow代码实例
# -*- coding: utf-8 -*- # Copyright 2015 Google Inc. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # ============================================================================== # 导入一些库 import collections import math import os import random import time import zipfile import numpy as np from six.moves import urllib from six.moves import xrange # pylint: disable=redefined-builtin import tensorflow as tf # 第一步: 下载数据. url = 'http://mattmahoney.net/dc/' def maybe_download(filename, expected_bytes): """如果文件存在, 下载文件, 并且保证文件大小正确""" if not os.path.exists(filename): filename, _ = urllib.request.urlretrieve(url + filename, filename) statinfo = os.stat(filename) if statinfo.st_size == expected_bytes: print('Found and verified', filename) else: print(statinfo.st_size) raise Exception( 'Failed to verify ' + filename + '. Can you get to it with a browser?') return filename filename = maybe_download('text8.zip', 31344016) # 把数据读入到一个列表中, 每个元素就是一个单词啦. def read_data(filename): """解压出的文件获取第一个文件, 作为单词列表""" with zipfile.ZipFile(filename) as f: data = f.read(f.namelist()[0]).split() return data #单词总数: 17005207 words = read_data(filename) print('Data size', len(words)) # 第二步: 构造字典, 把非常稀少的单词替换为"UNK"(未知的单词标记). vocabulary_size = 50000 def build_dataset(words): count = [['UNK', -1]] count.extend(collections.Counter(words).most_common(vocabulary_size - 1)) dictionary = dict() for word, _ in count: dictionary[word] = len(dictionary) data = list() unk_count = 0 for word in words: if word in dictionary: index = dictionary[word] else: index = 0 # dictionary['UNK'] unk_count += 1 data.append(index) count[0][1] = unk_count reverse_dictionary = dict(zip(dictionary.values(), dictionary.keys())) return data, count, dictionary, reverse_dictionary # data 是一个list, 按照文章的单词顺序记录了每个单词在我们字典dictionary中的index, 即出现频率排名 # count 是所有单词的计数dict # dictionary是每个单词的出现频率排名, key是单词, value是排名 # reverse_dictionary是dictionary的key-value颠倒, key是排名, value是单词 data, count, dictionary, reverse_dictionary = build_dataset(words) del words # Hint to reduce memory. print('Most common words (+UNK)', count[:5]) print('Sample data', data[:10], [reverse_dictionary[i] for i in data[:10]]) data_index = 0 # 第三步: 为skip-gram模型生成训练块的函数 def generate_batch(batch_size, num_skips, skip_window): global data_index assert batch_size % num_skips == 0 assert num_skips <= 2 * skip_window batch = np.ndarray(shape=(batch_size), dtype=np.int32) labels = np.ndarray(shape=(batch_size, 1), dtype=np.int32) span = 2 * skip_window + 1 buffer = collections.deque(maxlen=span) for _ in range(span): buffer.append(data[data_index]) data_index = (data_index + 1) % len(data) for i in range(batch_size // num_skips): target = skip_window # 在buffer中心的目标label targets_to_avoid = [skip_window] for j in range(num_skips): while target in targets_to_avoid: target = random.randint(0, span - 1) targets_to_avoid.append(target) batch[i * num_skips + j] = buffer[skip_window] labels[i * num_skips + j, 0] = buffer[target] buffer.append(data[data_index]) data_index = (data_index + 1) % len(data) return batch, labels batch, labels = generate_batch(batch_size=8, num_skips=2, skip_window=1) for i in range(8): print(batch[i], reverse_dictionary[batch[i]], '->', labels[i, 0], reverse_dictionary[labels[i, 0]]) # 第四步: 建立并训练skip-gram模型. batch_size = 128 embedding_size = 128 # embedding向量的维数, 即隐层维数 skip_window = 1 # 向左和向右考虑的单词数, 即向左向右仅考虑一个单词. num_skips = 2 # 可以重复使用输入去生成label的次数. # 我们随机生成集合抽样邻近单词, # 这里我们选那些出现频率比较高的单词 valid_size = 16 # 评估相似性的单词随机集合. valid_window = 100 # 在分布首部选择样本. valid_examples = np.random.choice(valid_window, valid_size, replace=False) num_sampled = 64 #错分的样本 graph = tf.Graph() with graph.as_default(): # 输入数据 train_inputs = tf.placeholder(tf.int32, shape=[batch_size]) train_labels = tf.placeholder(tf.int32, shape=[batch_size, 1]) valid_dataset = tf.constant(valid_examples, dtype=tf.int32) # 如果没有GPU,就用CPU的选项 with tf.device('/cpu:0'): # 在输入数据中寻找隐含层. embeddings = tf.Variable( tf.random_uniform([vocabulary_size, embedding_size], -1.0, 1.0)) embed = tf.nn.embedding_lookup(embeddings, train_inputs) # 为NCE 损失构造变量 nce_weights = tf.Variable( tf.truncated_normal([vocabulary_size, embedding_size], stddev=1.0 / math.sqrt(embedding_size))) nce_biases = tf.Variable(tf.zeros([vocabulary_size])) # 为训练输入计算平均NCE损失 # tf.nce计算损失时, 自动拿一个新的错分样本 loss = tf.reduce_mean( tf.nn.nce_loss(nce_weights, nce_biases, embed, train_labels, num_sampled, vocabulary_size)) # 学习率为1.0的SGD优化器 optimizer = tf.train.GradientDescentOptimizer(1.0).minimize(loss) # 为每个输入样本和所有隐含层计算cos相似度. norm = tf.sqrt(tf.reduce_sum(tf.square(embeddings), 1, keep_dims=True)) normalized_embeddings = embeddings / norm valid_embeddings = tf.nn.embedding_lookup( normalized_embeddings, valid_dataset) similarity = tf.matmul( valid_embeddings, normalized_embeddings, transpose_b=True) # Tensorboard作图用的Summarywriter loss_summary = tf.scalar_summary("loss", loss) train_summary_op = tf.merge_summary([loss_summary]) # 增加变量初始化 init = tf.initialize_all_variables() # 第五步: 开始训练!. num_steps = 100001 with tf.Session(graph=graph) as session: init.run() print("Initialized") # Directory in which to write summary information. # You can point TensorBoard to this directory via: # $ tensorboard --logdir=/tmp/word2vec_basic/summaries # Tensorflow assumes this directory already exists, so we need to create it. timestamp = str(int(time.time())) if not os.path.exists(os.path.join("/tmp/word2vec_basic", "summaries", timestamp)): os.makedirs(os.path.join("/tmp/word2vec_basic", "summaries", timestamp)) # 创建SummaryWriter train_summary_writer = tf.train.SummaryWriter( os.path.join( "/tmp/word2vec_basic", "summaries", timestamp), session.graph) average_loss = 0 for step in xrange(num_steps): batch_inputs, batch_labels = generate_batch( batch_size, num_skips, skip_window) feed_dict = {train_inputs: batch_inputs, train_labels: batch_labels} # We perform one update step by evaluating the optimizer op (including it # in the list of returned values for session.run() # Also evaluate the training summary op. _, loss_val, tsummary = session.run( [optimizer, loss, train_summary_op], feed_dict=feed_dict) average_loss += loss_val # Write the evaluated summary info to the SummaryWriter. This info will # then show up in the TensorBoard events. train_summary_writer.add_summary(tsummary, step) if step % 2000 == 0: if step > 0: average_loss /= 2000 # 平均损失是以往2000个输入的损失估计. print("Average loss at step ", step, ": ", average_loss) average_loss = 0 # 注意! 这很耗CPU(每经过500步就会慢下来大约20%) if step % 10000 == 0: sim = similarity.eval() for i in xrange(valid_size): valid_word = reverse_dictionary[valid_examples[i]] top_k = 8 # number of nearest neighbors nearest = (-sim[i, :]).argsort()[1:top_k + 1] log_str = "Nearest to %s:" % valid_word for k in xrange(top_k): close_word = reverse_dictionary[nearest[k]] log_str = "%s %s," % (log_str, close_word) print(log_str) final_embeddings = normalized_embeddings.eval() # 第六步: 可视化embeddings. def plot_with_labels(low_dim_embs, labels, filename='tsne.png'): assert low_dim_embs.shape[0] >= len(labels), "More labels than embeddings" plt.figure(figsize=(18, 18)) # in inches for i, label in enumerate(labels): x, y = low_dim_embs[i, :] plt.scatter(x, y) plt.annotate(label, xy=(x, y), xytext=(5, 2), textcoords='offset points', ha='right', va='bottom') plt.savefig(filename) try: from sklearn.manifold import TSNE import matplotlib.pyplot as plt tsne = TSNE(perplexity=30, n_components=2, init='pca', n_iter=5000) plot_only = 500 low_dim_embs = tsne.fit_transform(final_embeddings[:plot_only, :]) labels = [reverse_dictionary[i] for i in xrange(plot_only)] plot_with_labels(low_dim_embs, labels) except ImportError: print("Please install sklearn and matplotlib to visualize embeddings.")
参考文献
重要的Tensorflow资料:
- Tensorflow backgroud 是一个官方的Tensorflow动画教程非常棒:http://playground.tensorflow.org/
- TFLearn:一个深度学习的tensorflow上层API库。https://github.com/tflearn/tflearn
- 一些Tensorflow模型的实现: https://github.com/tensorflow/models
研讨会视频:
https://www.youtube.com/watch?v=GZBIPwdGtkk
研讨会PPT下载:
Loading...
The following two tabs change content below.
David 9
邮箱:yanchao727@gmail.com
微信: david9ml
Latest posts by David 9 (see all)
- 修订特征已经变得切实可行, “特征矫正工程”是否会成为潮流? - 27 3 月, 2024
- 量子计算系列#2 : 量子机器学习与量子深度学习补充资料,QML,QeML,QaML - 29 2 月, 2024
- “现象意识”#2:用白盒的视角研究意识和大脑,会是什么景象?微意识,主体感,超心智,意识中层理论 - 16 2 月, 2024
《Pycon 2016 tensorflow 研讨会总结 — tensorflow 手把手入门 #第二讲 word2vec》上有1条评论