Split (1)

train · 2.22k rows

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.

text stringlengths 35 1.54k	source stringclasses 1 value	page int64 1 800	book stringclasses 1 value
a model that can accommodate variable length inputs and variable length outputs. an rnn provides this ability. section describes several ways 10. 2. 4 of constructing an rnn that represents a conditional distribution over a sequence given some input, and section describes how to accomplish this conditioning 10. 4 when the input is a sequence. in all cases, one model ﬁrst reads the input sequence and emits a data structure that summarizes the input sequence. we call this summary the “ context ” c. the context c may be a list of vectors, or it may be a vector or tensor. the model that reads the input to produce c may be an rnn (, ; cho et al. 2014a sutskever 2014 jean 2014 et al., ; et al., ) or a convolutional network ( kalchbrenner and blunsom 2013, ). a second model, usually an rnn, then reads the context c and generates a sentence in the target language. this general idea of an encoder - decoder framework for machine translation is illustrated in ﬁgure. 12. 5 in order to generate an entire sentence conditioned on the source sentence, the model must have a way to represent the entire source sentence. earlier models	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	489	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
of an encoder - decoder framework for machine translation is illustrated in ﬁgure. 12. 5 in order to generate an entire sentence conditioned on the source sentence, the model must have a way to represent the entire source sentence. earlier models were only able to represent individual words or phrases. from a representation 474	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	489	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications learning point of view, it can be useful to learn a representation in which sentences that have the same meaning have similar representations regardless of whether they were written in the source language or the target language. this strategy was explored ﬁrst using a combination of convolutions and rnns ( kalchbrenner and blunsom 2013, ). later work introduced the use of an rnn for scoring proposed translations (, ) and for generating translated sentences ( cho et al. 2014a sutskever et al. et al., ). 2014 jean ( ) scaled these models to larger vocabularies. 2014 12. 4. 5. 1 using an attention mechanism and aligning pieces of data α ( t−1 ) α ( t−1 ) α ( ) t α ( ) t α ( + 1 ) t α ( + 1 ) t h ( t−1 ) h ( t−1 ) h ( ) t h ( ) t h ( + 1 ) t h ( + 1 ) t c × × × + figure 12. 6 : a modern attention mechanism, as introduced by ( ), is bahdanau et al. 2015 essentially a weighted average. a context vectorc is formed by taking a weighted average of feature vectors h	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	490	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
c × × × + figure 12. 6 : a modern attention mechanism, as introduced by ( ), is bahdanau et al. 2015 essentially a weighted average. a context vectorc is formed by taking a weighted average of feature vectors h ( ) t with weights α ( ) t. in some applications, the feature vectorsh are hidden units of a neural network, but they may also be raw input to the model. the weights α ( ) t are produced by the model itself. they are usually values in the interval [ 0, 1 ] and are intended to concentrate around just one h ( ) t so that the weighted average approximates reading that one speciﬁc time step precisely. the weightsα ( ) t are usually produced by applying a softmax function to relevance scores emitted by another portion of the model. the attention mechanism is more expensive computationally than directly indexing the desired h ( ) t, but direct indexing cannot be trained with gradient descent. the attention mechanism based on weighted averages is a smooth, [UNK] approximation that can be trained with existing optimization algorithms. using a ﬁxed - size representation to capture all the semantic details of a very long sentence of say 60 words is very [UNK]. it can be achieved by training a [UNK] large rn	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	490	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
a smooth, [UNK] approximation that can be trained with existing optimization algorithms. using a ﬁxed - size representation to capture all the semantic details of a very long sentence of say 60 words is very [UNK]. it can be achieved by training a [UNK] large rnn well enough and for long enough, as demonstrated by cho et al. ( ) and 2014a sutskever 2014 et al. ( ). however, a more [UNK] approach is to read the whole sentence or paragraph ( to get the context and the gist of what 475	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	490	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications is being expressed ), then produce the translated words one at a time, each time focusing on a [UNK] part of the input sentence in order to gather the semantic details that are required to produce the next output word. that is exactly the idea that ( ) ﬁrst introduced. the attention mechanism used bahdanau et al. 2015 to focus on speciﬁc parts of the input sequence at each time step is illustrated in ﬁgure. 12. 6 we can think of an attention - based system as having three components : 1. a process that “ reads ” raw data ( such as source words in a source sentence ), and converts them into distributed representations, with one feature vector associated with each word position. 2. a list of feature vectors storing the output of the reader. this can be understood as a “ ” containing a sequence of facts, which can be memory retrieved later, not necessarily in the same order, without having to visit all of them. 3. a process that “ ” the content of the memory to sequentially perform exploits a task, at each time step having the ability put attention on the content of one memory element ( or a few, with a [UNK] weight ). the third component generates the translated sentence. when words	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	491	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
the content of the memory to sequentially perform exploits a task, at each time step having the ability put attention on the content of one memory element ( or a few, with a [UNK] weight ). the third component generates the translated sentence. when words in a sentence written in one language are aligned with correspond - ing words in a translated sentence in another language, it becomes possible to relate the corresponding word embeddings. earlier work showed that one could learn a kind of translation matrix relating the word embeddings in one language with the word embeddings in another ( kocisky 2014 et al., ), yielding lower alignment error rates than traditional approaches based on the frequency counts in the phrase table. there is even earlier work on learning cross - lingual word vectors ( klementiev et al., 2012 ). many extensions to this approach are possible. for example, more [UNK] cross - lingual alignment (, ) allows training on larger datasets. gouws et al. 2014 12. 4. 6 historical perspective the idea of distributed representations for symbols was introduced by rumelhart et al. ( ) in one of the ﬁrst explorations of back - propagation, with symbols 1986a corresponding to the identity of family members	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	491	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
2014 12. 4. 6 historical perspective the idea of distributed representations for symbols was introduced by rumelhart et al. ( ) in one of the ﬁrst explorations of back - propagation, with symbols 1986a corresponding to the identity of family members and the neural network capturing the relationships between family members, with training examples forming triplets such as ( colin, mother, victoria ). the ﬁrst layer of the neural network learned a representation of each family member. for example, the features for colin 476	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	491	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications might represent which family tree colin was in, what branch of that tree he was in, what generation he was from, etc. one can think of the neural network as computing learned rules relating these attributes together in order to obtain the desired predictions. the model can then make predictions such as inferring who is the mother of colin. the idea of forming an embedding for a symbol was extended to the idea of an embedding for a word by deerwester 1990 et al. ( ). these embeddings were learned using the svd. later, embeddings would be learned by neural networks. the history of natural language processing is marked by transitions in the popularity of [UNK] ways of representing the input to the model. following this early work on symbols or words, some of the earliest applications of neural networks to nlp (, ; miikkulainen and dyer 1991 schmidhuber 1996, ) represented the input as a sequence of characters. bengio 2001 et al. ( ) returned the focus to modeling words and introduced neural language models, which produce interpretable word embeddings. these neural models have scaled up from deﬁning representations of a small set of symbols in the 1980s to millions of words	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	492	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
al. ( ) returned the focus to modeling words and introduced neural language models, which produce interpretable word embeddings. these neural models have scaled up from deﬁning representations of a small set of symbols in the 1980s to millions of words ( including proper nouns and misspellings ) in modern applications. this computational scaling [UNK] led to the invention of the techniques described above in section. 12. 4. 3 initially, the use of words as the fundamental units of language models yielded improved language modeling performance (, ). to this day, bengio et al. 2001 new techniques continually push both character - based models ( sutskever et al., 2011 ) and word - based models forward, with recent work (, ) even gillick et al. 2015 modeling individual bytes of unicode characters. the ideas behind neural language models have been extended into several natural language processing applications, such as parsing (,, ; henderson 2003 2004 collobert 2011, ), part - of - speech tagging, semantic role labeling, chunking, etc, sometimes using a single multi - task learning architecture ( collobert and weston, 2008a collobert 2011a ; et al., ) in which the word embeddings are shared across tasks. two	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	492	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
labeling, chunking, etc, sometimes using a single multi - task learning architecture ( collobert and weston, 2008a collobert 2011a ; et al., ) in which the word embeddings are shared across tasks. two - dimensional visualizations of embeddings became a popular tool for an - alyzing language models following the development of the t - sne dimensionality reduction algorithm ( van der maaten and hinton 2008, ) and its high - proﬁle appli - cation to visualization word embeddings by joseph turian in 2009. 477	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	492	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications 12. 5 other applications in this section we cover a few other types of applications of deep learning that are [UNK] from the standard object recognition, speech recognition and natural language processing tasks discussed above. part of this book will expand that iii scope even further to tasks that remain primarily research areas. 12. 5. 1 recommender systems one of the major families of applications of machine learning in the information technology sector is the ability to make recommendations of items to potential users or customers. two major types of applications can be distinguished : online advertising and item recommendations ( often these recommendations are still for the purpose of selling a product ). both rely on predicting the association between a user and an item, either to predict the probability of some action ( the user buying the product, or some proxy for this action ) or the expected gain ( which may depend on the value of the product ) if an ad is shown or a recommendation is made regarding that product to that user. the internet is currently ﬁnanced in great part by various forms of online advertising. there are major parts of the economy that rely on online shopping. companies including amazon and ebay use machine learning, including deep learning, for their product recommendations. sometimes, the items are not products that are actually for sale.	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	493	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
forms of online advertising. there are major parts of the economy that rely on online shopping. companies including amazon and ebay use machine learning, including deep learning, for their product recommendations. sometimes, the items are not products that are actually for sale. examples include selecting posts to display on social network news feeds, recommending movies to watch, recommending jokes, recommending advice from experts, matching players for video games, or matching people in dating services. often, this association problem is handled like a supervised learning problem : given some information about the item and about the user, predict the proxy of interest ( user clicks on ad, user enters a rating, user clicks on a “ like ” button, user buys product, user spends some amount of money on the product, user spends time visiting a page for the product, etc ). this often ends up being either a regression problem ( predicting some conditional expected value ) or a probabilistic classiﬁcation problem ( predicting the conditional probability of some discrete event ). the early work on recommender systems relied on minimal information as inputs for these predictions : the user id and the item id. in this context, the only way to generalize is to rely on the similarity between the patterns of values of the target variable for	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	493	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
early work on recommender systems relied on minimal information as inputs for these predictions : the user id and the item id. in this context, the only way to generalize is to rely on the similarity between the patterns of values of the target variable for [UNK] users or for [UNK] items. suppose that user 1 and user 2 both like items a, b and c. from this, we may infer that user 1 and 478	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	493	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications user 2 have similar tastes. if user 1 likes item d, then this should be a strong cue that user 2 will also like d. algorithms based on this principle come under the name of collaborative ﬁltering. both non - parametric approaches ( such as nearest - neighbor methods based on the estimated similarity between patterns of preferences ) and parametric methods are possible. parametric methods often rely on learning a distributed representation ( also called an embedding ) for each user and for each item. bilinear prediction of the target variable ( such as a rating ) is a simple parametric method that is highly successful and often found as a component of state - of - the - art systems. the prediction is obtained by the dot product between the user embedding and the item embedding ( possibly corrected by constants that depend only on either the user id or the item id ). let [UNK] be the matrix containing our predictions, a a matrix with user embeddings in its rows and b a matrix with item embeddings in its columns. let b and c be vectors that contain respectively a kind of bias for each user ( representing how grumpy or positive that user is in general ) and for each item ( representing its general	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	494	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
a matrix with item embeddings in its columns. let b and c be vectors that contain respectively a kind of bias for each user ( representing how grumpy or positive that user is in general ) and for each item ( representing its general popularity ). the bilinear prediction is thus obtained as follows : [UNK], i = bu + ci + j au, jbj, i. ( 12. 20 ) typically one wants to minimize the squared error between predicted ratings [UNK], i and actual ratings ru, i. user embeddings and item embeddings can then be conveniently visualized when they are ﬁrst reduced to a low dimension ( two or three ), or they can be used to compare users or items against each other, just like word embeddings. one way to obtain these embeddings is by performing a singular value decomposition of the matrix r of actual targets ( such as ratings ). this corresponds to factorizing r = udv ( or a normalized variant ) into the product of two factors, the lower rank matrices a = ud and b = v. one problem with the svd is that it treats the missing entries in an arbitrary way, as if they corresponded to a target value of 0.	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	494	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
variant ) into the product of two factors, the lower rank matrices a = ud and b = v. one problem with the svd is that it treats the missing entries in an arbitrary way, as if they corresponded to a target value of 0. instead we would like to avoid paying any cost for the predictions made on missing entries. fortunately, the sum of squared errors on the observed ratings can also be easily minimized by gradient - based optimization. the svd and the bilinear prediction of equation both 12. 20 performed very well in the competition for the netﬂix prize (, bennett and lanning 2007 ), aiming at predicting ratings for ﬁlms, based only on previous ratings by a large set of anonymous users. many machine learning experts participated in this competition, which took place between 2006 and 2009. it raised the level of research in recommender systems using advanced machine learning and yielded improvements in recommender systems. even though it did not win by itself, the simple bilinear prediction or svd was a component of the ensemble models 479	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	494	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications presented by most of the competitors, including the winners (, ; toscher et al. 2009 koren 2009, ). beyond these bilinear models with distributed representations, one of the ﬁrst uses of neural networks for collaborative ﬁltering is based on the rbm undirected probabilistic model ( salakhutdinov 2007 et al., ). rbms were an important element of the ensemble of methods that won the netﬂix competition ( toscher 2009 et al., ; koren 2009, ). more advanced variants on the idea of factorizing the ratings matrix have also been explored in the neural networks community ( salakhutdinov and mnih 2008, ). however, there is a basic limitation of collaborative ﬁltering systems : when a new item or a new user is introduced, its lack of rating history means that there is no way to evaluate its similarity with other items or users ( respectively ), or the degree of association between, say, that new user and existing items. this is called the problem of cold - start recommendations. a general way of solving the cold - start recommendation problem is to introduce extra information about the individual users and items. for example, this extra information could be	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	495	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
, that new user and existing items. this is called the problem of cold - start recommendations. a general way of solving the cold - start recommendation problem is to introduce extra information about the individual users and items. for example, this extra information could be user proﬁle information or features of each item. systems that use such information are called content - based recommender systems. the mapping from a rich set of user features or item features to an embedding can be learned through a deep learning architecture (, ; huang et al. 2013 elkahky 2015 et al., ). specialized deep learning architectures such as convolutional networks have also been applied to learn to extract features from rich content such as from musical audio tracks, for music recommendation ( van den oord 2013 et al., ). in that work, the convolutional net takes acoustic features as input and computes an embedding for the associated song. the dot product between this song embedding and the embedding for a user is then used to predict whether a user will listen to the song. 12. 5. 1. 1 exploration versus exploitation when making recommendations to users, an issue arises that goes beyond ordinary supervised learning and into the realm of reinforcement learning	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	495	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
##ding for a user is then used to predict whether a user will listen to the song. 12. 5. 1. 1 exploration versus exploitation when making recommendations to users, an issue arises that goes beyond ordinary supervised learning and into the realm of reinforcement learning. many recom - mendation problems are most accurately described theoretically as contextual bandits (, ;, ). the issue is that when we langford and zhang 2008 lu et al. 2010 use the recommendation system to collect data, we get a biased and incomplete view of the preferences of users : we only see the responses of users to the items they were recommended and not to the other items. in addition, in some cases we may not get any information on users for whom no recommendation has been made ( for example, with ad auctions, it may be that the price proposed for an 480	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	495	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications ad was below a minimum price threshold, or does not win the auction, so the ad is not shown at all ). more importantly, we get no information about what outcome would have resulted from recommending any of the other items. this would be like training a classiﬁer by picking one class [UNK] for each training example x ( typically the class with the highest probability according to the model ) and then only getting as feedback whether this was the correct class or not. clearly, each example conveys less information than in the supervised case where the true label y is directly accessible, so more examples are necessary. worse, if we are not careful, we could end up with a system that continues picking the wrong decisions even as more and more data is collected, because the correct decision initially had a very low probability : until the learner picks that correct decision, it does not learn about the correct decision. this is similar to the situation in reinforcement learning where only the reward for the selected action is observed. in general, reinforcement learning can involve a sequence of many actions and many rewards. the bandits scenario is a special case of reinforcement learning, in which the learner takes only a single action and receives a single reward. the bandit problem is easier in the	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	496	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
in general, reinforcement learning can involve a sequence of many actions and many rewards. the bandits scenario is a special case of reinforcement learning, in which the learner takes only a single action and receives a single reward. the bandit problem is easier in the sense that the learner knows which reward is associated with which action. in the general reinforcement learning scenario, a high reward or a low reward might have been caused by a recent action or by an action in the distant past. the term contextual bandits refers to the case where the action is taken in the context of some input variable that can inform the decision. for example, we at least know the user identity, and we want to pick an item. the mapping from context to action is also called a policy. the feedback loop between the learner and the data distribution ( which now depends on the actions of the learner ) is a central research issue in the reinforcement learning and bandits literature. reinforcement learning requires choosing a [UNK] exploration and exploitation. exploitation refers to taking actions that come from the current, best version of the learned policy — actions that we know will achieve a high reward. exploration refers to taking actions speciﬁcally in order to obtain more training data. if we know that given context x, action a gives	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	496	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
from the current, best version of the learned policy — actions that we know will achieve a high reward. exploration refers to taking actions speciﬁcally in order to obtain more training data. if we know that given context x, action a gives us a reward of 1, we do not know whether that is the best possible reward. we may want to exploit our current policy and continue taking action a in order to be relatively sure of obtaining a reward of 1. however, we may also want to explore by trying action a. we do not know what will happen if we try action a. we hope to get a reward of, but we 2 run the risk of getting a reward of. either way, we at least gain some knowledge. 0 exploration can be implemented in many ways, ranging from occasionally taking random actions intended to cover the entire space of possible actions, to model - based approaches that compute a choice of action based on its expected reward and the model ’ s amount of uncertainty about that reward. 481	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	496	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications many factors determine the extent to which we prefer exploration or exploitation. one of the most prominent factors is the time scale we are interested in. if the agent has only a short amount of time to accrue reward, then we prefer more exploitation. if the agent has a long time to accrue reward, then we begin with more exploration so that future actions can be planned more [UNK] with more knowledge. as time progresses and our learned policy improves, we move toward more exploitation. supervised learning has no [UNK] between exploration and exploitation because the supervision signal always speciﬁes which output is correct for each input. there is no need to try out [UNK] outputs to determine if one is better than the model ’ s current output — we always know that the label is the best output. another [UNK] arising in the context of reinforcement learning, besides the exploration - exploitation trade - [UNK], is the [UNK] of evaluating and comparing [UNK] policies. reinforcement learning involves interaction between the learner and the environment. this feedback loop means that it is not straightforward to evaluate the learner ’ s performance using a ﬁxed set of test set input values. the policy itself determines which inputs will be seen. ( ) present dudik et al. 2011 techniques for evaluating contextual bandits	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	497	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
that it is not straightforward to evaluate the learner ’ s performance using a ﬁxed set of test set input values. the policy itself determines which inputs will be seen. ( ) present dudik et al. 2011 techniques for evaluating contextual bandits. 12. 5. 2 knowledge representation, reasoning and question an - swering deep learning approaches have been very successful in language modeling, machine translation and natural language processing due to the use of embeddings for symbols (, ) and words ( rumelhart et al. 1986a deerwester 1990 bengio et al., ; et al., 2001 ). these embeddings represent semantic knowledge about individual words and concepts. a research frontier is to develop embeddings for phrases and for relations between words and facts. search engines already use machine learning for this purpose but much more remains to be done to improve these more advanced representations. 12. 5. 2. 1 knowledge, relations and question answering one interesting research direction is determining how distributed representations can be trained to capture the relations between two entities. these relations allow us to formalize facts about objects and how objects interact with each other. in mathematics, a binary relation is a set of ordered pairs of objects. pairs that are in the set are	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	497	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
be trained to capture the relations between two entities. these relations allow us to formalize facts about objects and how objects interact with each other. in mathematics, a binary relation is a set of ordered pairs of objects. pairs that are in the set are said to have the relation while those who are not in the set 482	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	497	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications do not. for example, we can deﬁne the relation “ is less than ” on the set of entities { 1, 2, 3 } by deﬁning the set of ordered pairs s = { ( 1, 2 ), ( 1, 3 ), ( 2, 3 ) }. once this relation is deﬁned, we can use it like a verb. because ( 1, 2 ) ∈s, we say that 1 is less than 2. because ( 2, 1 ) ∈s, we can not say that 2 is less than 1. of course, the entities that are related to one another need not be numbers. we could deﬁne a relation containing tuples like (, ). is _ a _ type _ of dog mammal in the context of ai, we think of a relation as a sentence in a syntactically simple and highly structured language. the relation plays the role of a verb, while two arguments to the relation play the role of its subject and object. these sentences take the form of a triplet of tokens ( subject verb object ),, ( 12. 21 ) with values ( entityi, relationj, entityk ). ( 12. 22 ) we can also deﬁ	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	498	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
and object. these sentences take the form of a triplet of tokens ( subject verb object ),, ( 12. 21 ) with values ( entityi, relationj, entityk ). ( 12. 22 ) we can also deﬁne an attribute, a concept analogous to a relation, but taking only one argument : ( entityi, attribute j ). ( 12. 23 ) for example, we could deﬁne the has _ fur attribute, and apply it to entities like dog. many applications require representing relations and reasoning about them. how should we best do this within the context of neural networks? machine learning models of course require training data. we can infer relations between entities from training datasets consisting of unstructured natural language. there are also structured databases that identify relations explicitly. a common structure for these databases is the relational database, which stores this same kind of information, albeit not formatted as three token sentences. when a database is intended to convey commonsense knowledge about everyday life or expert knowledge about an application area to an artiﬁcial intelligence system, we call the database a knowledge base. knowledge bases range from general ones like freebase, opencyc, wordnet, or wikibase, 1 etc. to more	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	498	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
expert knowledge about an application area to an artiﬁcial intelligence system, we call the database a knowledge base. knowledge bases range from general ones like freebase, opencyc, wordnet, or wikibase, 1 etc. to more specialized knowledge bases, like geneontology. 2 representations for entities and relations can be learned by considering each triplet in a knowledge base as a training example and maximizing a training objective that captures their joint distribution ( bordes et al., ). 2013a 1respectively available from these web sites : freebase. com, cyc. com / opencyc, wordnet. princeton. edu wikiba. se, 2geneontology. org 483	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	498	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications in addition to training data, we also need to deﬁne a model family to train. a common approach is to extend neural language models to model entities and relations. neural language models learn a vector that provides a distributed representation of each word. they also learn about interactions between words, such as which word is likely to come after a sequence of words, by learning functions of these vectors. we can extend this approach to entities and relations by learning an embedding vector for each relation. in fact, the parallel between modeling language and modeling knowledge encoded as relations is so close that researchers have trained representations of such entities by using both and knowledge bases natural language sentences (,, ; bordes et al. 2011 2012 wang 2014a et al., ) or combining data from multiple relational databases (, ). many bordes et al. 2013b possibilities exist for the particular parametrization associated with such a model. early work on learning about relations between entities (, paccanaro and hinton 2000 ) posited highly constrained parametric forms ( “ linear relational embeddings ” ), often using a [UNK] form of representation for the relation than for the entities. for example, paccanaro and hinton 2000 bordes 2011 ( ) and	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	499	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
posited highly constrained parametric forms ( “ linear relational embeddings ” ), often using a [UNK] form of representation for the relation than for the entities. for example, paccanaro and hinton 2000 bordes 2011 ( ) and et al. ( ) used vectors for entities and matrices for relations, with the idea that a relation acts like an operator on entities. alternatively, relations can be considered as any other entity ( bordes et al., ), allowing us to make statements about relations, but more ﬂexibility is 2012 put in the machinery that combines them in order to model their joint distribution. a practical short - term application of such models is link prediction : predict - ing missing arcs in the knowledge graph. this is a form of generalization to new facts, based on old facts. most of the knowledge bases that currently exist have been constructed through manual labor, which tends to leave many and probably the majority of true relations absent from the knowledge base. see wang et al. ( ), ( ) and ( ) for examples of such an 2014b lin et al. 2015 garcia - duran et al. 2015 application. evaluating the performance of a model on a link prediction task is [UNK] because we have only a dataset of positive examples (	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	499	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
( ) and ( ) for examples of such an 2014b lin et al. 2015 garcia - duran et al. 2015 application. evaluating the performance of a model on a link prediction task is [UNK] because we have only a dataset of positive examples ( facts that are known to be true ). if the model proposes a fact that is not in the dataset, we are unsure whether the model has made a mistake or discovered a new, previously unknown fact. the metrics are thus somewhat imprecise and are based on testing how the model ranks a held - out of set of known true positive facts compared to other facts that are less likely to be true. a common way to construct interesting examples that are probably negative ( facts that are probably false ) is to begin with a true fact and create corrupted versions of that fact, for example by replacing one entity in the relation with a [UNK] entity selected at random. the popular precision at 10 % metric counts how many times the model ranks a “ correct ” fact among the top 10 % of all corrupted versions of that fact. 484	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	499	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
chapter 12. applications another application of knowledge bases and distributed representations for them is word - sense disambiguation ( navigli and velardi 2005 bordes, ; et al., 2012 ), which is the task of deciding which of the senses of a word is the appropriate one, in some context. eventually, knowledge of relations combined with a reasoning process and understanding of natural language could allow us to build a general question answering system. a general question answering system must be able to process input information and remember important facts, organized in a way that enables it to retrieve and reason about them later. this remains a [UNK] open problem which can only be solved in restricted “ toy ” environments. currently, the best approach to remembering and retrieving speciﬁc declarative facts is to use an explicit memory mechanism, as described in section. memory networks were 10. 12 ﬁrst proposed to solve a toy question answering task ( weston 2014 kumar et al., ). et al. ( ) have proposed an extension that uses gru recurrent nets to read 2015 the input into the memory and to produce the answer given the contents of the memory. deep learning has been applied to many other applications besides the ones described here, and will surely be applied to	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	500	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
an extension that uses gru recurrent nets to read 2015 the input into the memory and to produce the answer given the contents of the memory. deep learning has been applied to many other applications besides the ones described here, and will surely be applied to even more after this writing. it would be impossible to describe anything remotely resembling a comprehensive coverage of such a topic. this survey provides a representative sample of what is possible as of this writing. this concludes part, which has described modern practices involving deep ii networks, comprising all of the most successful methods. generally speaking, these methods involve using the gradient of a cost function to ﬁnd the parameters of a model that approximates some desired function. with enough training data, this approach is extremely powerful. we now turn to part, in which we step into the iii territory of research — methods that are designed to work with less training data or to perform a greater variety of tasks, where the challenges are more [UNK] and not as close to being solved as the situations we have described so far. 485	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	500	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
part iii deep learning research 486	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	501	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
this part of the book describes the more ambitious and advanced approaches to deep learning, currently pursued by the research community. in the previous parts of the book, we have shown how to solve supervised learning problems — how to learn to map one vector to another, given enough examples of the mapping. not all problems we might want to solve fall into this category. we may wish to generate new examples, or determine how likely some point is, or handle missing values and take advantage of a large set of unlabeled examples or examples from related tasks. a shortcoming of the current state of the art for industrial applications is that our learning algorithms require large amounts of supervised data to achieve good accuracy. in this part of the book, we discuss some of the speculative approaches to reducing the amount of labeled data necessary for existing models to work well and be applicable across a broader range of tasks. accomplishing these goals usually requires some form of unsupervised or semi - supervised learning. many deep learning algorithms have been designed to tackle unsupervised learning problems, but none have truly solved the problem in the same way that deep learning has largely solved the supervised learning problem for a wide variety of tasks. in this part of the book, we describe the existing approaches to un	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	502	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
unsupervised learning problems, but none have truly solved the problem in the same way that deep learning has largely solved the supervised learning problem for a wide variety of tasks. in this part of the book, we describe the existing approaches to unsupervised learning and some of the popular thought about how we can make progress in this ﬁeld. a central cause of the [UNK] with unsupervised learning is the high di - mensionality of the random variables being modeled. this brings two distinct challenges : a statistical challenge and a computational challenge. the statistical challenge regards generalization : the number of conﬁgurations we may want to distinguish can grow exponentially with the number of dimensions of interest, and this quickly becomes much larger than the number of examples one can possibly have ( or use with bounded computational resources ). the computational challenge associated with high - dimensional distributions arises because many algorithms for learning or using a trained model ( especially those based on estimating an explicit probability function ) involve intractable computations that grow exponentially with the number of dimensions. with probabilistic models, this computational challenge arises from the need to perform intractable inference or simply from the need to normalize the distribution. • intractable inference : inference is discussed mostly in	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	502	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
that grow exponentially with the number of dimensions. with probabilistic models, this computational challenge arises from the need to perform intractable inference or simply from the need to normalize the distribution. • intractable inference : inference is discussed mostly in chapter. it regards 19 the question of guessing the probable values of some variables a, given other variables b, with respect to a model that captures the joint distribution over 487	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	502	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
a, b and c. in order to even compute such conditional probabilities one needs to sum over the values of the variables c, as well as compute a normalization constant which sums over the values of a and c. • intractable normalization constants ( the partition function ) : the partition function is discussed mostly in chapter. normalizing constants of proba - 18 bility functions come up in inference ( above ) as well as in learning. many probabilistic models involve such a normalizing constant. unfortunately, learning such a model often requires computing the gradient of the loga - rithm of the partition function with respect to the model parameters. that computation is generally as intractable as computing the partition function itself. monte carlo markov chain ( mcmc ) methods ( chapter ) are of - 17 ten used to deal with the partition function ( computing it or its gradient ). unfortunately, mcmc methods [UNK] when the modes of the model distribu - tion are numerous and well - separated, especially in high - dimensional spaces ( section ). 17. 5 one way to confront these intractable computations is to approximate them, and many approaches have been proposed as discussed in this third part of the book. another interesting way,	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	503	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)
- separated, especially in high - dimensional spaces ( section ). 17. 5 one way to confront these intractable computations is to approximate them, and many approaches have been proposed as discussed in this third part of the book. another interesting way, also discussed here, would be to avoid these intractable computations altogether by design, and methods that do not require such computations are thus very appealing. several generative models have been proposed in recent years, with that motivation. a wide variety of contemporary approaches to generative modeling are discussed in chapter. 20 part is the most important for a researcher — someone who wants to un - iii derstand the breadth of perspectives that have been brought to the ﬁeld of deep learning, and push the ﬁeld forward towards true artiﬁcial intelligence. 488	/home/ricoiban/GEMMA/mnlp_chatsplaining/RAG/Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org).pdf	503	Deep Learning by Ian Goodfellow, Yoshua Bengio, Aaron Courville (z-lib.org)