blocks|key|348436|text|以上都不是。默认情况下，它是随机初始化的embeddings_initializer='uniform'。|type|unstyled|depth|inlineStyleRanges|offset|length|style|CODE|entityRanges|data|348437|有一个关于如何使用现有嵌入的教程。|348438|entityMap|0|LINK|mutability|MUTABLE|url|https://blog.keras.io/using-pre-trained-word-embeddings-in-a-keras-model.html^0|K|W|0|E|2|0|0^^$0|@$1|2|3|4|5|6|7|P|8|@$9|Q|A|R|B|C]]|D|@]|E|$]]|$1|F|3|G|5|6|7|S|8|@]|D|@$9|T|A|U|1|V]]|E|$]]|$1|H|3|-4|5|6|7|W|8|@]|D|@]|E|$]]]|I|$J|$5|K|L|M|E|$N|O]]]]

None of the above. It is by default initialised randomly <code>embeddings_initializer='uniform'</code>.

There is a <a href="https://blog.keras.io/using-pre-trained-word-embeddings-in-a-keras-model.html" rel="nofollow noreferrer">tutorial</a> on how to use existing embeddings.

blocks|key|1192523|text|简单地说，和都不是。本质上，嵌入层(如Word2Vec+of+GloVe+)只是一个小的神经网络模块(通常是完全连接的层)，它将更高的稀疏维数投射到一个较低的n维向量中。|type|unstyled|depth|inlineStyleRanges|offset|length|style|BOLD|entityRanges|data|1192524|当您在Keras中插入一个新的随机嵌入层到您的神经网络中时，Keras将构造一个形状[input_dim,+output_dim]的密集可学习矩阵。|CODE|1192525|具体来说，假设您要插入一个嵌入层，将整数标量月份信息(12个唯一值)编码到大小为3的浮点向量中。|1192526|import+numpy+as+np
import+keras
from+keras.models+import+Sequential,+Model
from+keras.layers+import+Embedding,+Input
x+=+Input(shape=(1000,))+#+suppose+seq_len=1000
embedding+=+Embedding(12%2B1,+3,+input_length=1000)(x)
model+=+Model(inputs=x,+outputs=+embedding)+#+Functional+API
model.summary()|code-block|syntax|javascript|1192527|您的嵌入层将有如下摘要：|1192528|_________________________________________________________________
Layer+(type)+++++++++++++++++Output+Shape++++++++++++++Param+#+++
=================================================================
input_1+(InputLayer)+++++++++(None,+1000)++++++++++++++0+++++++++
_________________________________________________________________
embedding_1+(Embedding)++++++(None,+1000,+3)+++++++++++39++++++++
=================================================================
Total+params:+39
Trainable+params:+39
Non-trainable+params:+0
_________________________________________________________________|1192529|注意，可学习的参数是39+=+13*3+(+Keras需要%2B1来编码不属于12个唯一月份中任何一个的值--以防万一)。|1192530|还请注意，虽然嵌入的输入形状是形状(None,+1000)，嵌入的输出是形状(None,+1000,+3)。这意味着[13,+3]大小的非常小的稠密权重矩阵被应用于，每个的输入时间步骤为1000个。这意味着，每个月0-11的整数输入都将转换为大小为(3,)的浮点向量。|1192531|这也意味着，当您从最后一层反向传播到嵌入层时，1000个时间步骤嵌入输出中每一个的梯度也将(以time_distributed方式)流到尺寸为[13,3]的小神经网络权重(本质上是嵌入层)。|1192532|此外，请参阅嵌入层的官方Keras文档：https://keras.io/layers/embeddings/。|1192533|entityMap|0|LINK|mutability|MUTABLE|url|https://keras.io/layers/embeddings/^0|5|1|0|16|N|0|0|0|0|0|A|9|0|H|C|12|F|1M|7|2Z|4|3G|4|2A|3|0|1B|G|1Z|6|2I|2|0|K|Z|0|0^^$0|@$1|2|3|4|5|6|7|19|8|@$9|1A|A|1B|B|C]]|D|@]|E|$]]|$1|F|3|G|5|6|7|1C|8|@$9|1D|A|1E|B|H]]|D|@]|E|$]]|$1|I|3|J|5|6|7|1F|8|@]|D|@]|E|$]]|$1|K|3|L|5|M|7|1G|8|@]|D|@]|E|$N|O]]|$1|P|3|Q|5|6|7|1H|8|@]|D|@]|E|$]]|$1|R|3|S|5|M|7|1I|8|@]|D|@]|E|$N|O]]|$1|T|3|U|5|6|7|1J|8|@$9|1K|A|1L|B|H]]|D|@]|E|$]]|$1|V|3|W|5|6|7|1M|8|@$9|1N|A|1O|B|H]|$9|1P|A|1Q|B|H]|$9|1R|A|1S|B|H]|$9|1T|A|1U|B|H]|$9|1V|A|1W|B|H]|$9|1X|A|1Y|B|C]]|D|@]|E|$]]|$1|X|3|Y|5|6|7|1Z|8|@$9|20|A|21|B|H]|$9|22|A|23|B|H]|$9|24|A|25|B|C]]|D|@]|E|$]]|$1|Z|3|10|5|6|7|26|8|@]|D|@$9|27|A|28|1|29]]|E|$]]|$1|11|3|-4|5|6|7|2A|8|@]|D|@]|E|$]]]|12|$13|$5|14|15|16|E|$17|18]]]]

The short answer is neither. In essence, an embedding layer such as Word2Vec of GloVe is just a small neural network module (fully-connected layer usually) that projects higher, sparse dimensionality into a lower, n-dimensional vector.

When you insert a fresh random embedding layer in Keras into your neural network, Keras will construct a dense learnable matrix of shape <code>[input_dim, output_dim]</code>.

Concretely, let's say that you're inserting an Embedding layer to encode integer scalar month information (12 unique values) into a float vector of size 3. In Keras, you're going to declare your embedding as follows:

<pre><code>import numpy as np
import keras
from keras.models import Sequential, Model
from keras.layers import Embedding, Input
x = Input(shape=(1000,)) # suppose seq_len=1000
embedding = Embedding(12+1, 3, input_length=1000)(x)
model = Model(inputs=x, outputs= embedding) # Functional API
model.summary()
</code></pre>

Your embedding layer would have a summary as follows:

<pre><code>_________________________________________________________________
Layer (type) Output Shape Param # 
=================================================================
input_1 (InputLayer) (None, 1000) 0 
_________________________________________________________________
embedding_1 (Embedding) (None, 1000, 3) 39 
=================================================================
Total params: 39
Trainable params: 39
Non-trainable params: 0
_________________________________________________________________
</code></pre>

Notice that the learnable parameters are <code>39 = 13*3</code> (the +1 is needed by Keras to encode values that don't belong to any of the 12 unique months - just in case).

Also notice that while the input shape to embedding is shaped <code>(None, 1000)</code>, the output of the embedding is shaped <code>(None, 1000, 3)</code>. This means the very small dense weight matrix of size <code>[13, 3]</code> is applied to each of the 1000 input time-steps. Which means, every month integer input of <code>0-11</code> will be converted into a float vector of size <code>(3,)</code>.

This also means that when you do backpropagation from the final layer into the embedding layer, the gradient to each of the 1000 time-steps embedding output will also flow (in a <code>time_distributed</code> manner) to the small neural network weights (which is, essentially, the embedding layer) of size <code>[13,3]</code>.

Please also refer to official Keras documentation for Embedding layer: <a href="https://keras.io/layers/embeddings/" rel="noreferrer">https://keras.io/layers/embeddings/</a>.

Which algorithm is used for embedding in Keras built-in function?
Word2vec? Glove? Other?

<a href="https://keras.io/layers/embeddings/" rel="nofollow noreferrer">https://keras.io/layers/embeddings/</a>

Embedding in Keras

翻译质量差，导致语言生硬或混乱。

没有提供实际的解决方法或示例。

解答不清晰，无法理解或解决问题。

页面排版不美观，阅读体验差。

文章

问答

视频

学习中心

腾讯云实验室

直播

竞赛

腾讯云代码分析专区

腾讯iOA零信任安全管理系统专区

腾讯云架构师技术同盟交流圈

腾讯云数据库专区

腾讯云顾问专区

腾讯云原生专区

腾讯混元专区

腾讯云TCE专区

腾讯云Lighthouse专区

腾讯云HAI专区

腾讯云Edgeone专区

腾讯云存储专区

腾讯云智能专区

腾讯轻联专区 

腾讯云开发专区

TAPD专区

腾讯轻量云游戏服专区

腾讯云最具价值专家

腾讯云架构师技术同盟

腾讯云创作之星

腾讯云开发者先锋

腾讯云代码助手

云原生构建

TAPD 敏捷项目管理

Cloud Studio

SDK中心

API中心

命令行工具

涵盖代码开发、场景应用、自动测试全流程，助你从零构建专属AI助手

一站式MCP教程库，解锁AI应用新玩法

在Keras内建函数中使用哪种嵌入算法？Word2vec？手套？另一个？

问嵌入Keras
EN

Stack Overflow用户

社区

活动

圈层

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问嵌入KerasEN