In other words, is there a paper that describes the method of keras embedding layer? Is there a comparison between these methods (and other methods like Glove etc.)?
Solved – difference between keras embedding layer and word2vec
kerasnatural languageneural networksword embeddingsword2vec
Related Solutions
In fact, the output vectors are not computed from the input using any mathematical operation. Instead, each input integer is used as the index to access a table that contains all possible vectors. That is the reason why you need to specify the size of the vocabulary as the first argument (so the table can be initialized).
The most common application of this layer is for text processing. Let's see a simple example. Our training set consists only of two phrases:
Hope to see you soon
Nice to see you again
So we can encode these phrases by assigning each word a unique integer number (by order of appearance in our training dataset for example). Then our phrases could be rewritten as:
[0, 1, 2, 3, 4]
[5, 1, 2, 3, 6]
Now imagine we want to train a network whose first layer is an embedding layer. In this case, we should initialize it as follows:
Embedding(7, 2, input_length=5)
The first argument (7) is the number of distinct words in the training set. The second argument (2) indicates the size of the embedding vectors. The input_length argument, of course, determines the size of each input sequence.
Once the network has been trained, we can get the weights of the embedding layer, which in this case will be of size (7, 2) and can be thought as the table used to map integers to embedding vectors:
+------------+------------+
| index | Embedding |
+------------+------------+
| 0 | [1.2, 3.1] |
| 1 | [0.1, 4.2] |
| 2 | [1.0, 3.1] |
| 3 | [0.3, 2.1] |
| 4 | [2.2, 1.4] |
| 5 | [0.7, 1.7] |
| 6 | [4.1, 2.0] |
+------------+------------+
So according to these embeddings, our second training phrase will be represented as:
[[0.7, 1.7], [0.1, 4.2], [1.0, 3.1], [0.3, 2.1], [4.1, 2.0]]
It might seem counterintuitive at first, but the underlying automatic differentiation engines (e.g., Tensorflow or Theano) manage to optimize these vectors associated with each input integer just like any other parameter of your model.
For an intuition of how this table lookup is implemented as a mathematical operation which can be handled by the automatic differentiation engines, consider the embeddings table from the example as a (7, 2) matrix. Then, for a given word, you create a one-hot vector based on its index and multiply it by the embeddings matrix, effectively replicating a lookup. For instance, for the word "soon" the index is 4, and the one-hot vector is [0, 0, 0, 0, 1, 0, 0]. If you multiply this (1, 7) matrix by the (7, 2) embeddings matrix you get the desired two-dimensional embedding, which in this case is [2.2, 1.4].
It is also interesting to use the embeddings learned by other methods/people in different domains (see https://blog.keras.io/using-pre-trained-word-embeddings-in-a-keras-model.html) as done in [1].
[1] López-Sánchez, D., Herrero, J. R., Arrieta, A. G., & Corchado, J. M. Hybridizing metric learning and case-based reasoning for adaptable clickbait detection. Applied Intelligence, 1-16.
Embedding layers in Keras are trained just like any other layer in your network architecture: they are tuned to minimize the loss function by using the selected optimization method. The major difference with other layers, is that their output is not a mathematical function of the input. Instead the input to the layer is used to index a table with the embedding vectors [1]. However, the underlying automatic differentiation engine has no problem to optimize these vectors to minimize the loss function...
So, you cannot say that the Embedding layer in Keras is doing the same as word2vec [2]. Remember that word2vec refers to a very specific network setup which tries to learn an embedding which captures the semantics of words. With Keras's embedding layer, you are just trying to minimize the loss function, so if for instance you are working with a sentiment classification problem, the learned embedding will probably not capture complete word semantics but just their emotional polarity...
For example, the following image taken from [3] shows the embedding of three sentences with a Keras Embedding layer trained from scratch as part of a supervised network designed to detect clickbait headlines (left) and pre-trained word2vec embeddings (right). As you can see, word2vec embeddings reflect the semantic similarity between phrases b) and c). Conversely, the embeddings generated by Keras's Embedding layer might be useful for classification, but do not capture the semantical similarity of b) and c).
This explains why when you have a limited amount of training samples, it might be a good idea to initialize your Embedding layer with word2vec weights, so at least your model recognizes that "Alps" and "Himalaya" are similar things, even if they don't both occur in sentences of your training dataset.
[1] How does Keras 'Embedding' layer work?
[2] https://www.tensorflow.org/tutorials/word2vec
[3] https://link.springer.com/article/10.1007/s10489-017-1109-7
NOTE: Actually, the image shows the activations of the layer after the Embedding layer, but for the purpose of this example it does not matter... See more details in [3]
Best Answer
Embeddings (in general, not only in Keras) are methods for learning vector representations of categorical data. They are most commonly used for working with textual data. Word2vec and GloVe are two popular frameworks for learning word embeddings. What embeddings do, is they simply learn to map the one-hot encoded categorical variables to vectors of floating point numbers of smaller dimensionality then the input vectors. For example, one-hot vector representing a word from vocabulary of size 50 000 is mapped to real-valued vector of size 100. Then, the embeddings vector is used for whatever you want to use it as features.
The difference is how Word2vec is trained, as compared to the "usual" learned embeddings layers. Word2vec is trained to predict if word belongs to the context, given other words, e.g. to tell if "milk" is a likely word given the "The cat was drinking..." sentence begging. By doing so, we expect Word2vec to learn something about the language, as in the quote "You shall know a word by the company it keeps" by John Rupert Firth. Using the above example, Word2vec learns that "cat" is something that is likely to appear together with "milk", but also with "house", or "pet", so it is somehow similar to "dog". As a consequence, embeddings created by Word2vec, or similar models, learn to represent words with similar meanings using similar vectors.
On another hand, with embeddings learned as a layer of a neural network, the network may be trained to predict whatever you want. For example, you can train your network to predict sentiment of a text. In such case, the embeddings would learn features that are relevant for this particular problem. As a side effect, they can learn also some general things about the language, but the network is not optimized for such task. Using the "cat" example, embeddings trained for sentiment analysis may learn that "cat" and "dog" are similar, because people often say nice things about their pets.
In practical terms, you can use the pretrained Word2vec embeddings as features of any neural network (or other algorithm). They can give you advantage if your data is small, since the pretrained embeddings were trained on large volumes of text. On another hand, there are examples showing that learning the embeddings from your data, optimized for a particular problem, may be more efficient (Qi et al, 2018).
Qi, Y., Sachan, D. S., Felix, M., Padmanabhan, S. J., & Neubig, G. (2018). When and Why are Pre-trained Word Embeddings Useful for Neural Machine Translation? arXiv preprint arXiv:1804.06323.