results for au:Cho_K in:cs

- Osteoporosis is a public health problem characterized by increased fracture risk secondary to low bone mass and microarchitectural deterioration of bone tissue. Almost all fractures of the hip require hospitalization and major surgery. Early diagnosis of osteoporosis plays an important role in preventing osteoporotic fracture. Magnetic resonance imaging (MRI) has been successfully performed to image trabecular bone architecture in vivo proving itself as the practical imaging modality for bone quality assessment. However, segmentation of the whole proximal femur is required to measure bone quality and assess fracture risk precisely. Manual segmentation of the proximal femur is time-intensive, limiting the use of MRI measurements in the clinical practice. To overcome this bottleneck, robust automatic proximal femur segmentation method is required. In this paper, we propose to use deep convolutional neural networks (CNNs) for an automatic proximal femur segmentation using structural MR images. We constructed a dataset with 62 volumetric MR scans that are manually-segmented for proximal femur. We performed experiments using two different CNN architectures and achieved a high dice similarity score of 0.95.
- We propose a neural machine translation architecture that models the surrounding text in addition to the source sentence. These models lead to better performance, both in terms of general translation quality and pronoun prediction, when trained on small corpora, although this improvement largely disappears when trained with a larger corpus. We also discover that attention-based neural machine translation is well suited for pronoun prediction and compares favorably with other approaches that were specifically designed for this task.
- Apr 19 2017 cs.CL arXiv:1704.05179v1We publicly release a new large-scale dataset, called SearchQA, for machine comprehension, or question-answering. Unlike recently released datasets, such as DeepMind CNN/DailyMail and SQuAD, the proposed SearchQA was constructed to reflect a full pipeline of general question-answering. That is, we start not from an existing article and generate a question-answer pair, but start from an existing question-answer pair, crawled from J! Archive, and augment it with text snippets retrieved by Google. Following this approach, we built SearchQA, which consists of more than 140k question-answer pairs with each pair having 49.6 snippets on average. Each question-answer-context tuple of the SearchQA comes with additional meta-data such as the snippet's URL, which we believe will be valuable resources for future research. We conduct human evaluation as well as test two baseline methods, one simple word selection and the other deep learning based, on the SearchQA. We show that there is a meaningful gap between the human and machine performances. This suggests that the proposed dataset could well serve as a benchmark for question-answering.
- Apr 18 2017 cs.IR arXiv:1704.04572v1Search engines play an important role in our everyday lives by assisting us in finding the information we need. When we input a complex query, however, results are often far from satisfactory. In this work, we introduce a query reformulation system based on a neural network that rewrites a query to maximize the number of relevant documents returned. We train this neural network with reinforcement learning. The actions correspond to selecting terms to build a reformulated query, and the reward is the document recall. We evaluate our approach on three datasets against strong baselines and show a relative improvement of 5-20% in terms of recall. Furthermore, we present a simple method to estimate a conservative upper-bound performance of a model in a particular environment and verify that there is still large room for improvements.
- In this paper, we present a transfer learning approach for music classification and regression tasks. We propose to use a pretrained convnet feature, a concatenated feature vector using activations of feature maps of multiple layers in a trained convolutional network. We show that how this convnet feature can serve as a general-purpose music representation. In the experiment, a convnet is trained for music tagging and then transferred for many music-related classification and regression tasks as well as an audio-related classification task. In experiments, the convnet feature outperforms the baseline MFCC feature in all tasks and many reported approaches of aggregating MFCCs and low- and high-level music features.
- Recent advances in deep learning for object recognition in natural images has prompted a surge of interest in applying a similar set of techniques to medical images. Most of the initial attempts largely focused on replacing the input to such a deep convolutional neural network from a natural image to a medical image. This, however, does not take into consideration the fundamental differences between these two types of data. More specifically, detection or recognition of an anomaly in medical images depends significantly on fine details, unlike object recognition in natural images where coarser, more global structures matter more. This difference makes it inadequate to use the existing deep convolutional neural networks architectures, which were developed for natural images, because they rely on heavily downsampling an image to a much lower resolution to reduce the memory requirements. This hides details necessary to make accurate predictions for medical images. Furthermore, a single exam in medical imaging often comes with a set of different views which must be seamlessly fused in order to reach a correct conclusion. In our work, we propose to use a multi-view deep convolutional neural network that handles a set of more than one high-resolution medical image. We evaluate this network on large-scale mammography-based breast cancer screening (BI-RADS prediction) using 103 thousand images. We focus on investigating the impact of training set sizes and image sizes on the prediction accuracy. Our results highlight that performance clearly increases with the size of training set, and that the best performance can only be achieved using the images in the original resolution. This suggests the future direction of medical imaging research using deep neural networks is to utilize as much data as possible with the least amount of potentially harmful preprocessing.
- Mar 14 2017 cs.CL arXiv:1703.04357v1We present Nematus, a toolkit for Neural Machine Translation. The toolkit prioritizes high translation accuracy, usability, and extensibility. Nematus has been used to build top-performing submissions to shared translation tasks at WMT and IWSLT, and has been used to train systems for production environments.
- Mar 07 2017 cs.NI arXiv:1703.02005v1In the mid-90's, it was shown that the statistics of aggregated time series from Internet traffic departed from those of traditional short range dependent models, and were instead characterized by asymptotic self-similarity. Following this seminal contribution, over the years, many studies have investigated the existence and form of scaling in Internet traffic. This contribution aims first at presenting a methodology, combining multiscale analysis (wavelet and wavelet leaders) and random projections (or sketches), permitting a precise, efficient and robust characterization of scaling which is capable of seeing through non-stationary anomalies. Second, we apply the methodology to a data set spanning an unusually long period: 14 years, from the MAWI traffic archive, thereby allowing an in-depth longitudinal analysis of the form, nature and evolutions of scaling in Internet traffic, as well as network mechanisms producing them. We also study a separate 3-day long trace to obtain complementary insight into intra-day behavior. We find that a biscaling (two ranges of independent scaling phenomena) regime is systematically observed: long-range dependence over the large scales, and multifractal-like scaling over the fine scales. We quantify the actual scaling ranges precisely, verify to high accuracy the expected relationship between the long range dependent parameter and the heavy tail parameter of the flow size distribution, and relate fine scale multifractal scaling to typical IP packet inter-arrival and to round-trip time distributions.
- We introduce a novel approach to training generative adversarial networks, where we train a generator to match a target distribution that converges to the data distribution at the limit of a perfect discriminator. This objective can be interpreted as training a generator to produce samples that lie on the decision boundary of a current discriminator in training at each update, and we call a GAN trained using this algorithm a boundary-seeking GAN (BS-GAN). This approach can be used to train a generator with discrete output when the generator outputs a parametric conditional distribution. We demonstrate the effectiveness of the proposed algorithm with discrete image data. In contrary to the proposed algorithm, we observe that the recently proposed Gumbel-Softmax technique for re-parametrizing the discrete variables does not work for training a GAN with discrete data. Finally, we notice that the proposed boundary-seeking algorithm works even with continuous variables, and demonstrate its effectiveness with two widely used image data sets, SVHN and CelebA.
- Feb 14 2017 cs.CL arXiv:1702.03525v2There has been relatively little attention to incorporating linguistic prior to neural machine translation. Much of the previous work was further constrained to considering linguistic prior on the source side. In this paper, we propose a hybrid model, called NMT+RNNG, that learns to parse and translate by combining the recurrent neural network grammar into the attention-based neural machine translation. Our approach encourages the neural machine translation model to incorporate linguistic prior during training, and lets it translate on its own afterward. Extensive experiments with four language pairs show the effectiveness of the proposed NMT+RNNG.
- Recent research in neural machine translation has largely focused on two aspects; neural network architectures and end-to-end learning algorithms. The problem of decoding, however, has received relatively little attention from the research community. In this paper, we solely focus on the problem of decoding given a trained neural machine translation model. Instead of trying to build a new decoding algorithm for any specific decoding objective, we propose the idea of trainable decoding algorithm in which we train a decoding algorithm to find a translation that maximizes an arbitrary decoding objective. More specifically, we design an actor that observes and manipulates the hidden state of the neural machine translation decoder and propose to train it using a variant of deterministic policy gradient. We extensively evaluate the proposed algorithm using four language pairs and two decoding objectives and show that we can indeed train a trainable greedy decoder that generates a better translation (in terms of a target decoding objective) with minimal computational overhead.
- Latent representation learned from multi-layered neural networks via hierarchical feature abstraction enables recent success of deep learning. Under the deep learning framework, generalization performance highly depends on the learned latent representation which is obtained from an appropriate training scenario with a task-specific objective on a designed network model. In this work, we propose a novel latent space modeling method to learn better latent representation. We designed a neural network model based on the assumption that good base representation can be attained by maximizing the total correlation between the input, latent, and output variables. From the base model, we introduce a semantic noise modeling method which enables class-conditional perturbation on latent space to enhance the representational power of learned latent feature. During training, latent vector representation can be stochastically perturbed by a modeled class-conditional additive noise while maintaining its original semantic feature. It implicitly brings the effect of semantic augmentation on the latent space. The proposed model can be easily learned by back-propagation with common gradient-based optimization algorithms. Experimental results show that the proposed method helps to achieve performance benefits against various previous approaches. We also provide the empirical analyses for the proposed class-conditional perturbation process including t-SNE visualization.
- Most existing machine translation systems operate at the level of words, relying on explicit segmentation to extract tokens. We introduce a neural machine translation (NMT) model that maps a source character sequence to a target character sequence without any segmentation. We employ a character-level convolutional network with max-pooling at the encoder to reduce the length of source representation, allowing the model to be trained at a speed comparable to subword-level models while capturing local regularities. Our character-to-character model outperforms a recently proposed baseline with a subword-level encoder on WMT'15 DE-EN and CS-EN, and gives comparable performance on FI-EN and RU-EN. We then demonstrate that it is possible to share a single character-level encoder across multiple languages by training a model on a many-to-one translation task. In this multilingual setting, the character-level encoder significantly outperforms the subword-level encoder on all the language pairs. We observe that on CS-EN, FI-EN and RU-EN, the quality of the multilingual character-level translation even surpasses the models specifically trained on that language pair alone, both in terms of BLEU score and human judgment.
- Translating in real-time, a.k.a. simultaneous translation, outputs translation words before the input sentence ends, which is a challenging problem for conventional machine translation methods. We propose a neural machine translation (NMT) framework for simultaneous translation in which an agent learns to make decisions on when to translate from the interaction with a pre-trained NMT environment. To trade off quality and delay, we extensively explore various targets for delay and design a method for beam-search applicable in the simultaneous MT setting. Experiments against state-of-the-art baselines on two language pairs demonstrate the efficacy of the proposed framework both quantitatively and qualitatively.
- We introduce a convolutional recurrent neural network (CRNN) for music tagging. CRNNs take advantage of convolutional neural networks (CNNs) for local feature extraction and recurrent neural networks for temporal summarisation of the extracted features. We compare CRNN with three CNN structures that have been used for music tagging while controlling the number of parameters with respect to their performance and training time per sample. Overall, we found that CRNNs show a strong performance with respect to the number of parameter and training time, indicating the effectiveness of its hybrid structure in music feature extraction and feature summarisation.
- Descriptions are often provided along with recommendations to help users' discovery. Recommending automatically generated music playlists (e.g. personalised playlists) introduces the problem of generating descriptions. In this paper, we propose a method for generating music playlist descriptions, which is called as music captioning. In the proposed method, audio content analysis and natural language processing are adopted to utilise the information of each track.
- Jul 05 2016 cs.CL arXiv:1607.00578v1We first observe a potential weakness of continuous vector representations of symbols in neural machine translation. That is, the continuous vector representation, or a word embedding vector, of a symbol encodes multiple dimensions of similarity, equivalent to encoding more than one meaning of the word. This has the consequence that the encoder and decoder recurrent networks in neural machine translation need to spend substantial amount of their capacity in disambiguating source and target words based on the context which is defined by a source sentence. Based on this observation, in this paper we propose to contextualize the word embedding vectors using a nonlinear bag-of-words representation of the source sentence. Additionally, we propose to represent special tokens (such as numbers, proper nouns and acronyms) with typed symbols to facilitate translating those words that are not well-suited to be translated via continuous vectors. The experiments on En-Fr and En-De reveal that the proposed approaches of contextualization and symbolization improves the translation quality of neural machine translation systems significantly.
- We extend neural Turing machine (NTM) model into a dynamic neural Turing machine (D-NTM) by introducing a trainable memory addressing scheme. This addressing scheme maintains for each memory cell two separate vectors, content and address vectors. This allows the D-NTM to learn a wide variety of location-based addressing strategies including both linear and nonlinear ones. We implement the D-NTM with both continuous, differentiable and discrete, non-differentiable read/write mechanisms. We investigate the mechanisms and effects of learning to read and write into a memory through experiments on Facebook bAbI tasks using both a feedforward and GRUcontroller. The D-NTM is evaluated on a set of Facebook bAbI tasks and shown to outperform NTM and LSTM baselines. We have done extensive analysis of our model and different variations of NTM on bAbI task. We also provide further experimental results on sequential pMNIST, Stanford Natural Language Inference, associative recall and copy tasks.
- Jun 17 2016 physics.soc-ph cs.DL arXiv:1606.04972v1Statistics of article page views is useful for measuring the impact of individual articles. Analyzing the temporal evolution of article page views, we find that article page views usually decay over time after reaching a peak, especially exhibiting relaxation with nonexponentiality. This finding suggests that relaxation in article page views resembles physical aging as frequently found in complex systems.
- Jun 16 2016 cs.CL arXiv:1606.04754v1Interlingua based Machine Translation (MT) aims to encode multiple languages into a common linguistic representation and then decode sentences in multiple target languages from this representation. In this work we explore this idea in the context of neural encoder decoder architectures, albeit on a smaller scale and without MT as the end goal. Specifically, we consider the case of three languages or modalities X, Z and Y wherein we are interested in generating sequences in Y starting from information available in X. However, there is no parallel training data available between X and Y but, training data is available between X & Z and Z & Y (as is often the case in many real world applications). Z thus acts as a pivot/bridge. An obvious solution, which is perhaps less elegant but works very well in practice is to train a two stage model which first converts from X to Z and then from Z to Y. Instead we explore an interlingua inspired solution which jointly learns to do the following (i) encode X and Z to a common representation and (ii) decode Y from this common representation. We evaluate our model on two tasks: (i) bridge transliteration and (ii) bridge captioning. We report promising results in both these applications and believe that this is a right step towards truly interlingua inspired encoder decoder architectures.
- Jun 15 2016 cs.CL arXiv:1606.04164v1In this paper, we propose a novel finetuning algorithm for the recently introduced multi-way, mulitlingual neural machine translate that enables zero-resource machine translation. When used together with novel many-to-one translation strategies, we empirically show that this finetuning algorithm allows the multi-way, multilingual model to translate a zero-resource language pair (1) as well as a single-pair neural translation model trained with up to 1M direct parallel sentences of the same language pair and (2) better than pivot-based translation strategy, while keeping only one additional copy of attention-related parameters.
- Jun 09 2016 cs.CL arXiv:1606.02680v1Neural machine translation has become a major alternative to widely used phrase-based statistical machine translation. We notice however that much of research on neural machine translation has focused on European languages despite its language agnostic nature. In this paper, we apply neural machine translation to the task of Arabic translation (Ar<->En) and compare it against a standard phrase-based translation system. We run extensive comparison using various configurations in preprocessing Arabic script and show that the phrase-based and neural translation systems perform comparably to each other and that proper preprocessing of Arabic script has a similar effect on both of the systems. We however observe that the neural machine translation significantly outperform the phrase-based system on an out-of-domain test set, making it attractive for real-world deployment.
- Jun 08 2016 cs.CL arXiv:1606.02012v1We investigate the potential of attention-based neural machine translation in simultaneous translation. We introduce a novel decoding algorithm, called simultaneous greedy decoding, that allows an existing neural machine translation model to begin translating before a full source sentence is received. This approach is unique from previous works on simultaneous translation in that segmentation and translation are done jointly to maximize the translation quality and that translating each segment is strongly conditioned on all the previous segments. This paper presents a first step toward building a full simultaneous translation system based on neural machine translation.
- Multivariate time series data in practical applications, such as health care, geoscience, and biology, are characterized by a variety of missing values. In time series prediction and other related tasks, it has been noted that missing values and their missing patterns are often correlated with the target labels, a.k.a., informative missingness. There is very limited work on exploiting the missing patterns for effective imputation and improving prediction performance. In this paper, we develop novel deep learning models, namely GRU-D, as one of the early attempts. GRU-D is based on Gated Recurrent Unit (GRU), a state-of-the-art recurrent neural network. It takes two representations of missing patterns, i.e., masking and time interval, and effectively incorporates them into a deep model architecture so that it not only captures the long-term temporal dependencies in time series, but also utilizes the missing patterns to achieve better prediction results. Experiments of time series classification tasks on real-world clinical datasets (MIMIC-III, PhysioNet) and synthetic datasets demonstrate that our models achieve state-of-the-art performance and provides useful insights for better understanding and utilization of missing values in time series analysis.
- Jun 07 2016 cs.CL arXiv:1606.01700v2We introduce a recurrent neural network language model (RNN-LM) with long short-term memory (LSTM) units that utilizes both character-level and word-level inputs. Our model has a gate that adaptively finds the optimal mixture of the character-level and word-level inputs. The gate creates the final vector representation of a word by combining two distinct representations of the word. The character-level inputs are converted into vector representations of words using a bidirectional LSTM. The word-level inputs are projected into another high-dimensional space by a word lookup table. The final vector representations of words are used in the LSTM language model which predicts the next word given all the preceding words. Our model with the gating mechanism effectively utilizes the character-level inputs for rare and out-of-vocabulary words and outperforms word-level language models on several English corpora.
- One way to approach end-to-end autonomous driving is to learn a policy function that maps from a sensory input, such as an image frame from a front-facing camera, to a driving action, by imitating an expert driver, or a reference policy. This can be done by supervised learning, where a policy function is tuned to minimize the difference between the predicted and ground-truth actions. A policy function trained in this way however is known to suffer from unexpected behaviours due to the mismatch between the states reachable by the reference policy and trained policy functions. More advanced algorithms for imitation learning, such as DAgger, addresses this issue by iteratively collecting training examples from both reference and trained policies. These algorithms often requires a large number of queries to a reference policy, which is undesirable as the reference policy is often expensive. In this paper, we propose an extension of the DAgger, called SafeDAgger, that is query-efficient and more suitable for end-to-end autonomous driving. We evaluate the proposed SafeDAgger in a car racing simulator and show that it indeed requires less queries to a reference policy. We observe a significant speed up in convergence, which we conjecture to be due to the effect of automated curriculum learning.
- Recent advances in conditional recurrent language modelling have mainly focused on network architectures (e.g., attention mechanism), learning algorithms (e.g., scheduled sampling and sequence-level training) and novel applications (e.g., image/video description generation, speech recognition, etc.) On the other hand, we notice that decoding algorithms/strategies have not been investigated as much, and it has become standard to use greedy or beam search. In this paper, we propose a novel decoding strategy motivated by an earlier observation that nonlinear hidden layers of a deep neural network stretch the data manifold. The proposed strategy is embarrassingly parallelizable without any communication overhead, while improving an existing decoding algorithm. We extensively evaluate it with attention-based neural machine translation on the task of En->Cz translation.
- Theano is a Python library that allows to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently. Since its introduction, it has been one of the most used CPU and GPU mathematical compilers - especially in the machine learning community - and has shown steady performance improvements. Theano is being actively and continuously developed since 2008, multiple frameworks have been built on top of it and it has been used to produce many state-of-the-art machine learning models. The present article is structured as follows. Section I provides an overview of the Theano software and its community. Section II presents the principal features of Theano and how to use them, and compares them with other similar projects. Section III focuses on recently-introduced functionalities and improvements. Section IV compares the performance of Theano against Torch7 and TensorFlow on several machine learning models. Section V discusses current limitations of Theano and potential ways of improving it.
- The existing machine translation systems, whether phrase-based or neural, have relied almost exclusively on word-level modelling with explicit segmentation. In this paper, we ask a fundamental question: can neural machine translation generate a character sequence without any explicit segmentation? To answer this question, we evaluate an attention-based encoder-decoder with a subword-level encoder and a character-level decoder on four language pairs--En-Cs, En-De, En-Ru and En-Fi-- using the parallel corpora from WMT'15. Our experiments show that the models with a character-level decoder outperform the ones with a subword-level decoder on all of the four language pairs. Furthermore, the ensembles of neural models with a character-level decoder outperform the state-of-the-art non-neural machine translation systems on En-Cs, En-De and En-Fi and perform comparably on En-Ru.
- We present a model of Selinger and Valiron's quantum lambda calculus based on von Neumann algebras, and show that the model is adequate with respect to the operational semantics.
- Unsupervised methods for learning distributed representations of words are ubiquitous in today's NLP research, but far less is known about the best ways to learn distributed phrase or sentence representations from unlabelled data. This paper is a systematic comparison of models that learn such representations. We find that the optimal approach depends critically on the intended application. Deeper, more complex models are preferable for representations to be used in supervised systems, but shallow log-linear models work best for building representation spaces that can be decoded with simple spatial distance metrics. We also propose two new unsupervised representation-learning objectives designed to optimise the trade-off between training time, domain portability and performance.
- Feb 09 2016 cs.AI arXiv:1602.02261v2We propose a goal-driven web navigation as a benchmark task for evaluating an agent with abilities to understand natural language and plan on partially observed environments. In this challenging task, an agent navigates through a website, which is represented as a graph consisting of web pages as nodes and hyperlinks as directed edges, to find a web page in which a query appears. The agent is required to have sophisticated high-level reasoning based on natural languages and efficient sequential decision-making capability to succeed. We release a software tool, called WebNav, that automatically transforms a website into this goal-driven web navigation task, and as an example, we make WikiNav, a dataset constructed from the English Wikipedia. We extensively evaluate different variants of neural net based artificial agents on WikiNav and observe that the proposed goal-driven web navigation well reflects the advances in models, making it a suitable benchmark for evaluating future progress. Furthermore, we extend the WikiNav with question-answer pairs from Jeopardy! and test the proposed agent based on recurrent neural networks against strong inverted index based search engines. The artificial agents trained on WikiNav outperforms the engined based approaches, demonstrating the capability of the proposed goal-driven navigation as a good proxy for measuring the progress in real-world tasks such as focused crawling and question-answering.
- Feb 02 2016 cs.CL arXiv:1602.00367v1Document classification tasks were primarily tackled at word level. Recent research that works with character-level inputs shows several benefits over word-level approaches such as natural incorporation of morphemes and better handling of rare words. We propose a neural network architecture that utilizes both convolution and recurrent layers to efficiently encode character inputs. We validate the proposed model on eight large scale document classification tasks and compare with character-level convolution-only models. It achieves comparable performances with much less parameters.
- We propose multi-way, multilingual neural machine translation. The proposed approach enables a single neural translation model to translate between multiple languages, with a number of parameters that grows only linearly with the number of languages. This is made possible by having a single attention mechanism that is shared across all language pairs. We train the proposed multi-way, multilingual model on ten language pairs from WMT'15 simultaneously and observe clear performance improvements over models trained on only one language pair. In particular, we observe that the proposed model significantly improves the translation quality of low-resource language pairs.
- Effectus theory is a new branch of categorical logic that aims to capture the essentials of quantum logic, with probabilistic and Boolean logic as special cases. Predicates in effectus theory are not subobjects having a Heyting algebra structure, like in topos theory, but `characteristic' functions, forming effect algebras. Such effect algebras are algebraic models of quantitative logic, in which double negation holds. Effects in quantum theory and fuzzy predicates in probability theory form examples of effect algebras. This text is an account of the basics of effectus theory. It includes the fundamental duality between states and effects, with the associated Born rule for validity of an effect (predicate) in a particular state. A basic result says that effectuses can be described equivalently in both `total' and `partial' form. So-called `commutative' and `Boolean' effectuses are distinguished, for probabilistic and classical models. It is shown how these Boolean effectuses are essentially extensive categories. A large part of the theory is devoted to the logical notions of comprehension and quotient, which are described abstractly as right adjoint to truth, and as left adjoint to falisity, respectively. It is illustrated how comprehension and quotients are closely related to measurement. The paper closes with a section on `non-commutative' effectus theory, where the appropriate formalisation is not entirely clear yet.
- This is a lecture note for the course DS-GA 3001 <Natural Language Understanding with Distributed Representation> at the Center for Data Science , New York University in Fall, 2015. As the name of the course suggests, this lecture note introduces readers to a neural network based approach to natural language understanding/processing. In order to make it as self-contained as possible, I spend much time on describing basics of machine learning and neural networks, only after which how they are used for natural languages is introduced. On the language front, I almost solely focus on language modelling and machine translation, two of which I personally find most fascinating and most fundamental to natural language understanding.
- We propose a structured prediction architecture, which exploits the local generic features extracted by Convolutional Neural Networks and the capacity of Recurrent Neural Networks (RNN) to retrieve distant dependencies. The proposed architecture, called ReSeg, is based on the recently introduced ReNet model for image classification. We modify and extend it to perform the more challenging task of semantic segmentation. Each ReNet layer is composed of four RNN that sweep the image horizontally and vertically in both directions, encoding patches or activations, and providing relevant global information. Moreover, ReNet layers are stacked on top of pre-trained convolutional layers, benefiting from generic local features. Upsampling layers follow ReNet layers to recover the original image resolution in the final predictions. The proposed ReSeg architecture is efficient, flexible and suitable for a variety of semantic segmentation tasks. We evaluate ReSeg on several widely-used semantic segmentation datasets: Weizmann Horse, Oxford Flower, and CamVid; achieving state-of-the-art performance. Results show that ReSeg can act as a suitable architecture for semantic segmentation tasks, and may have further applications in other structured prediction problems. The source code and model hyperparameters are available on https://github.com/fvisin/reseg.
- In this paper, we propose and study a novel visual object tracking approach based on convolutional networks and recurrent networks. The proposed approach is distinct from the existing approaches to visual object tracking, such as filtering-based ones and tracking-by-detection ones, in the sense that the tracking system is explicitly trained off-line to track anonymous objects in a noisy environment. The proposed visual tracking model is end-to-end trainable, minimizing any adversarial effect from mismatches in object representation and between the true underlying dynamics and learning dynamics. We empirically show that the proposed tracking approach works well in various scenarios by generating artificial video sequences with varying conditions; the number of objects, amount of noise and the match between the training shapes and test shapes.
- Variational methods that rely on a recognition network to approximate the posterior of directed graphical models offer better inference and learning than previous methods. Recent advances that exploit the capacity and flexibility in this approach have expanded what kinds of models can be trained. However, as a proposal for the posterior, the capacity of the recognition network is limited, which can constrain the representational power of the generative model and increase the variance of Monte Carlo estimates. To address these issues, we introduce an iterative refinement procedure for improving the approximate posterior of the recognition network and show that training with the refined posterior is competitive with state-of-the-art methods. The advantages of refinement are further evident in an increased effective sample size, which implies a lower variance of gradient estimates.
- Recently there has been growing interest in building active visual object recognizers, as opposed to the usual passive recognizers which classifies a given static image into a predefined set of object categories. In this paper we propose to generalize these recently proposed end-to-end active visual recognizers into a controller-recognizer framework. A model in the controller-recognizer framework consists of a controller, which interfaces with an external manipulator, and a recognizer which classifies the visual input adjusted by the manipulator. We describe two most recently proposed controller-recognizer models: recurrent attention model and spatial transformer network as representative examples of controller-recognizer models. Based on this description we observe that most existing end-to-end controller-recognizers tightly, or completely, couple a controller and recognizer. We ask a question whether this tight coupling is necessary, and try to answer this empirically by building a controller-recognizer model with a decoupled controller and recognizer. Our experiments revealed that it is not always necessary to tightly couple them and that by decoupling a controller and recognizer, there is a possibility of building a generic controller that is pretrained and works together with any subsequent recognizer.
- The task of associating images and videos with a natural language description has attracted a great amount of attention recently. Rapid progress has been made in terms of both developing novel algorithms and releasing new datasets. Indeed, the state-of-the-art results on some of the standard datasets have been pushed into the regime where it has become more and more difficult to make significant improvements. Instead of proposing new models, this work investigates the possibility of empirically establishing performance upper bounds on various visual captioning datasets without extra data labelling effort or human evaluation. In particular, it is assumed that visual captioning is decomposed into two steps: from visual inputs to visual concepts, and from visual concepts to natural language descriptions. One would be able to obtain an upper bound when assuming the first step is perfect and only requiring training a conditional language model for the second step. We demonstrate the construction of such bounds on MS-COCO, YouTube2Text and LSMDC (a combination of M-VAD and MPII-MD). Surprisingly, despite of the imperfect process we used for visual concept extraction in the first step and the simplicity of the language model for the second step, we show that current state-of-the-art models fall short when being compared with the learned upper bounds. Furthermore, with such a bound, we quantify several important factors concerning image and video captioning: the number of visual concepts captured by different models, the trade-off between the amount of visual elements captured and their accuracy, and the intrinsic difficulty and blessing of different datasets.
- Nov 13 2015 cs.CL arXiv:1511.03729v2In this work, we propose a novel method to incorporate corpus-level discourse information into language modelling. We call this larger-context language model. We introduce a late fusion approach to a recurrent language model based on long short-term memory units (LSTM), which helps the LSTM unit keep intra-sentence dependencies and inter-sentence dependencies separate from each other. Through the evaluation on three corpora (IMDB, BBC, and PennTree Bank), we demon- strate that the proposed model improves perplexity significantly. In the experi- ments, we evaluate the proposed approach while varying the number of context sentences and observe that the proposed late fusion is superior to the usual way of incorporating additional inputs to the LSTM. By analyzing the trained larger- context language model, we discover that content words, including nouns, adjec- tives and verbs, benefit most from an increasing number of context sentences. This analysis suggests that larger-context language model improves the unconditional language model by capturing the theme of a document better and more easily.
- Nov 06 2015 cs.LO arXiv:1511.01570v1Quotients and comprehension are fundamental mathematical constructions that can be described via adjunctions in categorical logic. This paper reveals that quotients and comprehension are related to measurement, not only in quantum logic, but also in probabilistic and classical logic. This relation is presented by a long series of examples, some of them easy, and some also highly non-trivial (esp. for von Neumann algebras). We have not yet identified a unifying theory. Nevertheless, the paper contributes towards such a theory by introducing the new quotient-and-comprehension perspective on measurement instruments, and by describing the examples on which such a theory should be built.
- This paper uncovers the fundamental relationship between total and partial computation in the form of an equivalence of certain categories. This equivalence involves on the one hand effectuses, which are categories for total computation, introduced by Jacobs for the study of quantum/effect logic. On the other hand, it involves what we call FinPACs with effects; they are finitely partially additive categories equipped with effect algebra structures, serving as categories for partial computation. It turns out that the Kleisli category of the lift monad (-)+1 on an effectus is always a FinPAC with effects, and this construction gives rise to the equivalence. Additionally, state-and-effect triangles over FinPACs with effects are presented.
- Whereas deep neural networks were first mostly used for classification tasks, they are rapidly expanding in the realm of structured output problems, where the observed target is composed of multiple random variables that have a rich joint distribution, given the input. We focus in this paper on the case where the input also has a rich structure and the input and output structures are somehow related. We describe systems that learn to attend to different places in the input, for each element of the output, for a variety of tasks: machine translation, image caption generation, video clip description and speech recognition. All these systems are based on a shared set of building blocks: gated recurrent neural networks and convolutional neural networks, along with trained attention mechanisms. We report on experimental results with these systems, showing impressively good performance and the advantage of the attention mechanism.
- Recurrent sequence generators conditioned on input data through an attention mechanism have recently shown very good performance on a range of tasks in- cluding machine translation, handwriting synthesis and image caption gen- eration. We extend the attention-mechanism with features needed for speech recognition. We show that while an adaptation of the model used for machine translation in reaches a competitive 18.7% phoneme error rate (PER) on the TIMIT phoneme recognition task, it can only be applied to utterances which are roughly as long as the ones it was trained on. We offer a qualitative explanation of this failure and propose a novel and generic method of adding location-awareness to the attention mechanism to alleviate this issue. The new method yields a model that is robust to long inputs and achieves 18% PER in single utterances and 20% in 10-times longer (repeated) utterances. Finally, we propose a change to the at- tention mechanism that prevents it from concentrating too much on single frames, which further reduces PER to 17.6% level.
- May 05 2015 cs.CV arXiv:1505.00393v3In this paper, we propose a deep neural network architecture for object recognition based on recurrent neural networks. The proposed network, called ReNet, replaces the ubiquitous convolution+pooling layer of the deep convolutional neural network with four recurrent neural networks that sweep horizontally and vertically in both directions across the image. We evaluate the proposed ReNet on three widely-used benchmark datasets; MNIST, CIFAR-10 and SVHN. The result suggests that ReNet is a viable alternative to the deep convolutional neural network, and that further investigation is needed.
- Apr 03 2015 cs.CL arXiv:1504.00548v4Distributional models that learn rich semantic word representations are a success story of recent NLP research. However, developing models that learn useful representations of phrases and sentences has proved far harder. We propose using the definitions found in everyday dictionaries as a means of bridging this gap between lexical and phrasal semantics. Neural language embedding models can be effectively trained to map dictionary definitions (phrases) to (lexical) representations of the words defined by those definitions. We present two applications of these architectures: "reverse dictionaries" that return the name of a concept given a definition or description and general-knowledge crossword question answerers. On both tasks, neural language embedding models trained on definitions from a handful of freely-available lexical resources perform as well or better than existing commercial systems that rely on significant task-specific engineering. The results highlight the effectiveness of both neural embedding architectures and definition-based training for developing models that understand phrases and sentences.
- Mar 13 2015 cs.CL arXiv:1503.03535v2Recent work on end-to-end neural network-based architectures for machine translation has shown promising results for En-Fr and En-De translation. Arguably, one of the major factors behind this success has been the availability of high quality parallel corpora. In this work, we investigate how to leverage abundant monolingual corpora for neural machine translation. Compared to a phrase-based and hierarchical baseline, we obtain up to $1.96$ BLEU improvement on the low-resource language pair Turkish-English, and $1.59$ BLEU on the focused domain task of Chinese-English chat messages. While our method was initially targeted toward such tasks with less parallel data, we show that it also extends to high resource languages such as Cs-En and De-En where we obtain an improvement of $0.39$ and $0.47$ BLEU scores over the neural machine translation baselines, respectively.
- Recent progress in using recurrent neural networks (RNNs) for image description has motivated the exploration of their application for video description. However, while images are static, working with videos requires modeling their dynamic temporal structure and then properly integrating that information into a natural language description. In this context, we propose an approach that successfully takes into account both the local and global temporal structure of videos to produce descriptions. First, our approach incorporates a spatial temporal 3-D convolutional neural network (3-D CNN) representation of the short temporal dynamics. The 3-D CNN representation is trained on video action recognition tasks, so as to produce a representation that is tuned to human motion and behavior. Second we propose a temporal attention mechanism that allows to go beyond local temporal modeling and learns to automatically select the most relevant temporal segments given the text-generating RNN. Our approach exceeds the current state-of-art for both BLEU and METEOR metrics on the Youtube2Text dataset. We also present results on a new, larger and more challenging dataset of paired video and natural language descriptions.
- Inspired by recent work in machine translation and object detection, we introduce an attention based model that automatically learns to describe the content of images. We describe how we can train this model in a deterministic manner using standard backpropagation techniques and stochastically by maximizing a variational lower bound. We also show through visualization how the model is able to automatically learn to fix its gaze on salient objects while generating the corresponding words in the output sequence. We validate the use of attention with state-of-the-art performance on three benchmark datasets: Flickr8k, Flickr30k and MS COCO.
- In this work, we propose a novel recurrent neural network (RNN) architecture. The proposed RNN, gated-feedback RNN (GF-RNN), extends the existing approach of stacking multiple recurrent layers by allowing and controlling signals flowing from upper recurrent layers to lower layers using a global gating unit for each pair of layers. The recurrent signals exchanged between layers are gated adaptively based on the previous hidden states and the current input. We evaluated the proposed GF-RNN with different types of recurrent units, such as tanh, long short-term memory and gated recurrent units, on the tasks of character-level language modeling and Python program evaluation. Our empirical evaluation of different RNN units, revealed that in both tasks, the GF-RNN outperforms the conventional approaches to build deep stacked RNNs. We suggest that the improvement arises because the GF-RNN can adaptively assign different layers to different timescales and layer-to-layer interactions (including the top-down ones which are not usually present in a stacked RNN) by learning to gate these interactions.
- This paper presents a novel semantics for a quantum programming language by operator algebras, which are known to give a formulation for quantum theory that is alternative to the one by Hilbert spaces. We show that the opposite category of the category of W*-algebras and normal completely positive subunital maps is an elementary quantum flow chart category in the sense of Selinger. As a consequence, it gives a denotational semantics for Selinger's first-order functional quantum programming language QPL. The use of operator algebras allows us to accommodate infinite structures and to handle classical and quantum computations in a unified way.
- Dec 22 2014 cs.CL arXiv:1412.6448v4Neural language models learn word representations, or embeddings, that capture rich linguistic and conceptual information. Here we investigate the embeddings learned by neural machine translation models, a recently-developed class of neural language model. We show that embeddings from translation models outperform those learned by monolingual models at tasks that require knowledge of both conceptual similarity and lexical-syntactic role. We further show that these effects hold when translating from both English to French and English to German, and argue that the desirable properties of translation embeddings should emerge largely independently of the source and target languages. Finally, we apply a new method for training neural translation models with very large vocabularies, and show that this vocabulary expansion algorithm results in minimal degradation of embedding quality. Our embedding spaces can be queried in an online demo and downloaded from our web page. Overall, our analyses indicate that translation-based embeddings should be used in applications that require concepts to be organised according to similarity and/or lexical function, while monolingual embeddings are better suited to modelling (nonspecific) inter-word relatedness.
- Dec 18 2014 cs.DC arXiv:1412.5557v2This is the final report on reproducibility@xsede, a one-day workshop held in conjunction with XSEDE14, the annual conference of the Extreme Science and Engineering Discovery Environment (XSEDE). The workshop's discussion-oriented agenda focused on reproducibility in large-scale computational research. Two important themes capture the spirit of the workshop submissions and discussions: (1) organizational stakeholders, especially supercomputer centers, are in a unique position to promote, enable, and support reproducible research; and (2) individual researchers should conduct each experiment as though someone will replicate that experiment. Participants documented numerous issues, questions, technologies, practices, and potentially promising initiatives emerging from the discussion, but also highlighted four areas of particular interest to XSEDE: (1) documentation and training that promotes reproducible research; (2) system-level tools that provide build- and run-time information at the level of the individual job; (3) the need to model best practices in research collaborations involving XSEDE staff; and (4) continued work on gateways and related technologies. In addition, an intriguing question emerged from the day's interactions: would there be value in establishing an annual award for excellence in reproducible research?
- In this paper we compare different types of recurrent units in recurrent neural networks (RNNs). Especially, we focus on more sophisticated units that implement a gating mechanism, such as a long short-term memory (LSTM) unit and a recently proposed gated recurrent unit (GRU). We evaluate these recurrent units on the tasks of polyphonic music modeling and speech signal modeling. Our experiments revealed that these advanced recurrent units are indeed better than more traditional recurrent units such as tanh units. Also, we found GRU to be comparable to LSTM.
- Dec 08 2014 cs.CL arXiv:1412.2007v2Neural machine translation, a recently proposed approach to machine translation based purely on neural networks, has shown promising results compared to the existing approaches such as phrase-based statistical machine translation. Despite its recent success, neural machine translation has its limitation in handling a larger vocabulary, as training complexity as well as decoding complexity increase proportionally to the number of target words. In this paper, we propose a method that allows us to use a very large target vocabulary without increasing training complexity, based on importance sampling. We show that decoding can be efficiently done even with the model having a very large target vocabulary by selecting only a small subset of the whole target vocabulary. The models trained by the proposed approach are empirically found to outperform the baseline models with a small vocabulary as well as the LSTM-based neural machine translation models. Furthermore, when we use the ensemble of a few models with very large target vocabularies, we achieve the state-of-the-art translation performance (measured by BLEU) on the English->German translation and almost as high performance as state-of-the-art English->French translation system.
- We replace the Hidden Markov Model (HMM) which is traditionally used in in continuous speech recognition with a bi-directional recurrent neural network encoder coupled to a recurrent neural network decoder that directly emits a stream of phonemes. The alignment between the input and output sequences is established using an attention mechanism: the decoder emits each symbol based on a context created with a subset of input symbols elected by the attention mechanism. We report initial results demonstrating that this new approach achieves phoneme error rates that are comparable to the state-of-the-art HMM-based decoders, on the TIMIT dataset.
- Oct 06 2014 cs.CL arXiv:1410.0718v2Neural language models learn word representations that capture rich linguistic and conceptual information. Here we investigate the embeddings learned by neural machine translation models. We show that translation-based embeddings outperform those learned by cutting-edge monolingual models at single-language tasks requiring knowledge of conceptual similarity and/or syntactic role. The findings suggest that, while monolingual models learn information about how concepts are related, neural-translation models better capture their true ontological status.
- The authors of (Cho et al., 2014a) have shown that the recently introduced neural network translation systems suffer from a significant drop in translation quality when translating long sentences, unlike existing phrase-based translation systems. In this paper, we propose a way to address this issue by automatically segmenting an input sentence into phrases that can be easily translated by the neural network translation model. Once each segment has been independently translated by the neural machine translation model, the translated clauses are concatenated to form a final translation. Empirical results show a significant improvement in translation quality for long sentences.
- Neural machine translation is a relatively new approach to statistical machine translation based purely on neural networks. The neural machine translation models often consist of an encoder and a decoder. The encoder extracts a fixed-length representation from a variable-length input sentence, and the decoder generates a correct translation from this representation. In this paper, we focus on analyzing the properties of the neural machine translation using two models; RNN Encoder--Decoder and a newly proposed gated recursive convolutional neural network. We show that the neural machine translation performs relatively well on short sentences without unknown words, but its performance degrades rapidly as the length of the sentence and the number of unknown words increase. Furthermore, we find that the proposed gated recursive convolutional network learns a grammatical structure of a sentence automatically.
- Neural Autoregressive Distribution Estimators (NADEs) have recently been shown as successful alternatives for modeling high dimensional multimodal distributions. One issue associated with NADEs is that they rely on a particular order of factorization for $P(\mathbf{x})$. This issue has been recently addressed by a variant of NADE called Orderless NADEs and its deeper version, Deep Orderless NADE. Orderless NADEs are trained based on a criterion that stochastically maximizes $P(\mathbf{x})$ with all possible orders of factorizations. Unfortunately, ancestral sampling from deep NADE is very expensive, corresponding to running through a neural net separately predicting each of the visible variables given some others. This work makes a connection between this criterion and the training criterion for Generative Stochastic Networks (GSNs). It shows that training NADEs in this way also trains a GSN, which defines a Markov chain associated with the NADE model. Based on this connection, we show an alternative way to sample from a trained Orderless NADE that allows to trade-off computing time and quality of the samples: a 3 to 10-fold speedup (taking into account the waste due to correlations between consecutive samples of the chain) can be obtained without noticeably reducing the quality of the samples. This is achieved using a novel sampling procedure for GSNs called annealed GSN sampling, similar to tempering methods that combines fast mixing (obtained thanks to steps at high noise levels) with accurate samples (obtained thanks to steps at low noise levels).
- Neural machine translation is a recently proposed approach to machine translation. Unlike the traditional statistical machine translation, the neural machine translation aims at building a single neural network that can be jointly tuned to maximize the translation performance. The models proposed recently for neural machine translation often belong to a family of encoder-decoders and consists of an encoder that encodes a source sentence into a fixed-length vector from which a decoder generates a translation. In this paper, we conjecture that the use of a fixed-length vector is a bottleneck in improving the performance of this basic encoder-decoder architecture, and propose to extend this by allowing a model to automatically (soft-)search for parts of a source sentence that are relevant to predicting a target word, without having to form these parts as a hard segment explicitly. With this new approach, we achieve a translation performance comparable to the existing state-of-the-art phrase-based system on the task of English-to-French translation. Furthermore, qualitative analysis reveals that the (soft-)alignments found by the model agree well with our intuition.
- Many state-of-the-art results obtained with deep networks are achieved with the largest models that could be trained, and if more computation power was available, we might be able to exploit much larger datasets in order to improve generalization ability. Whereas in learning algorithms such as decision trees the ratio of capacity (e.g., the number of parameters) to computation is very favorable (up to exponentially more parameters than computation), the ratio is essentially 1 for deep neural networks. Conditional computation has been proposed as a way to increase the capacity of a deep neural network without increasing the amount of computation required, by activating some parameters and computation "on-demand", on a per-example basis. In this note, we propose a novel parametrization of weight matrices in neural networks which has the potential to increase up to exponentially the ratio of the number of parameters to computation. The proposed approach is based on turning on some parameters (weight matrices) when specific bit patterns of hidden unit activations are obtained. In order to better control for the overfitting that might result, we propose a parametrization that is tree-structured, where each node of the tree corresponds to a prefix of a sequence of sign bits, or gating units, associated with hidden units.
- The Boolean satisfiability (SAT) problem is the first known example of an NP-complete problem, and thousands of NP-compete problems have been identified by reducing the SAT to the problems. Researchers have tried to find a definite mathematical expression that distinguishes among NL-complete, P-complete, and NP-complete problems such as 2-SAT, Horn-SAT, and 3-SAT. In this paper, we introduce the natural number system hidden inside the SAT structure. We reduce a SAT instance to an integer-programming instance. Then, we focus on the distance from an integral point to the facets of the projected polytope. We newly define a dominant variable, decision chain, and chain coupler as a novel element of a Boolean formula. From the analysis of the SAT structure using the elements, we show that the coefficients of the normal vector of the facet can be expressed with the natural number system of which the exponent is exponential in the input size. Furthermore, we prove that an integral point, which is not contained in the solution region, can locate exponentially near the projected polytope by the number system. Finally, we show that the number system is not formed in 2-SAT, but partially formed in Horn-SAT according to the feasible value of a dominant variable, and always formed in k-SAT (k>2) regardless of the feasible value of a dominant variable. Two questions, NL =? P and P =? NP, have been open problems for several decades. This study presents a definite supporting evidence for the conjecture that NL is a proper subset of P and P is a proper subset of NP, and a new solving direction for the P versus NP problem.
- A central challenge to many fields of science and engineering involves minimizing non-convex error functions over continuous, high dimensional spaces. Gradient descent or quasi-Newton methods are almost ubiquitously used to perform such minimizations, and it is often thought that a main source of difficulty for these local methods to find the global minimum is the proliferation of local minima with much higher error than the global minimum. Here we argue, based on results from statistical physics, random matrix theory, neural network theory, and empirical evidence, that a deeper and more profound difficulty originates from the proliferation of saddle points, not local minima, especially in high dimensional problems of practical interest. Such saddle points are surrounded by high error plateaus that can dramatically slow down learning, and give the illusory impression of the existence of a local minimum. Motivated by these arguments, we propose a new approach to second-order optimization, the saddle-free Newton method, that can rapidly escape high dimensional saddle points, unlike gradient descent and quasi-Newton methods. We apply this algorithm to deep or recurrent neural network training, and provide numerical evidence for its superior optimization performance.
- Training of the neural autoregressive density estimator (NADE) can be viewed as doing one step of probabilistic inference on missing values in data. We propose a new model that extends this inference scheme to multiple steps, arguing that it is easier to learn to improve a reconstruction in $k$ steps rather than to learn to reconstruct in a single inference step. The proposed model is an unsupervised building block for deep learning that combines the desirable properties of NADE and multi-predictive training: (1) Its test likelihood can be computed analytically, (2) it is easy to generate independent samples from it, and (3) it uses an inference engine that is a superset of variational inference for Boltzmann machines. The proposed NADE-k is competitive with the state-of-the-art in density estimation on the two datasets tested.
- In this paper, we propose a novel neural network model called RNN Encoder-Decoder that consists of two recurrent neural networks (RNN). One RNN encodes a sequence of symbols into a fixed-length vector representation, and the other decodes the representation into another sequence of symbols. The encoder and decoder of the proposed model are jointly trained to maximize the conditional probability of a target sequence given a source sequence. The performance of a statistical machine translation system is empirically found to improve by using the conditional probabilities of phrase pairs computed by the RNN Encoder-Decoder as an additional feature in the existing log-linear model. Qualitatively, we show that the proposed model learns a semantically and syntactically meaningful representation of linguistic phrases.
- We study the complexity of functions computable by deep feedforward neural networks with piecewise linear activations in terms of the symmetries and the number of linear regions that they have. Deep networks are able to sequentially map portions of each layer's input-space to the same output. In this way, deep models compute functions that react equally to complicated patterns of different inputs. The compositional structure of these functions enables them to re-use pieces of computation exponentially often in terms of the network's depth. This paper investigates the complexity of such compositional maps and contributes new theoretical results regarding the advantage of depth for neural networks with piecewise linear activation functions. In particular, our analysis is not specific to a single family of models, and as an example, we employ it for rectifier and maxout networks. We improve complexity bounds from pre-existing work and investigate the behavior of units in higher layers.
- In this paper, we explore different ways to extend a recurrent neural network (RNN) to a \textitdeep RNN. We start by arguing that the concept of depth in an RNN is not as clear as it is in feedforward neural networks. By carefully analyzing and understanding the architecture of an RNN, however, we find three points of an RNN which may be made deeper; (1) input-to-hidden function, (2) hidden-to-hidden transition and (3) hidden-to-output function. Based on this observation, we propose two novel architectures of a deep RNN which are orthogonal to an earlier attempt of stacking multiple recurrent layers to build a deep RNN (Schmidhuber, 1992; El Hihi and Bengio, 1996). We provide an alternative interpretation of these deep RNNs using a novel framework based on neural operators. The proposed deep RNNs are empirically evaluated on the tasks of polyphonic music prediction and language modeling. The experimental result supports our claim that the proposed deep RNNs benefit from the depth and outperform the conventional, shallow RNNs.