We address the question of characterizing and finding optimal representations for supervised learning. Traditionally, this question has been tackled using the Information Bottleneck, which compresses the inputs while retaining information about the targets, in a decoder-agnostic fashion. In machine learning, however, our goal is not compression but rather generalization, which is intimately linked to the predictive family or decoder of interest (e.g. linear classifier). We propose the Decodable Information Bottleneck (DIB) that considers information retention and compression from the perspective of the desired predictive family. As a result, DIB gives rise to representations that are optimal in terms of expected test performance and can be estimated with guarantees. Empirically, we show that the framework can be used to enforce a small generalization gap on downstream classifiers and to predict the generalization ability of neural networks.
Recent successes in deep generative modeling have led to significant advances in natural language generation (NLG). Incorporating entities into neural generation models has demonstrated great improvements by assisting to infer the summary topic and to generate coherent content. To enhance the role of entity in NLG, in this paper, we aim to model the entity type in the decoding phase to generate contextual words accurately. We develop a novel NLG model to produce a target sequence based on a given list of entities. Our model has a multi-step decoder that injects the entity types into the process of entity mention generation. Experiments on two public news datasets demonstrate type injection performs better than existing type embedding concatenation baselines.
This paper estimates free energy, average mutual information, and minimum mean square error (MMSE) of a linear model under two assumptions: (1) the source is generated by a Markov chain, (2) the source is generated via a hidden Markov model. Our estimates are based on the replica method in statistical physics. We show that under the posterior mean estimator, the linear model with Markov sources or hidden Markov sources is decoupled into single-input AWGN channels with state information available at both encoder and decoder where the state distribution follows the left Perron-Frobenius eigenvector with unit Manhattan norm of the stochastic matrix of Markov chains. Numerical results show that the free energies and MSEs obtained via the replica method are closely approximate to their counterparts achieved by the Metropolis-Hastings algorithm or some well-known approximate message passing algorithms in the research literature.
Drug discovery using deep learning has attracted a lot of attention of late as it has obvious advantages like higher efficiency, less manual guessing and faster process time. In this paper, we present a novel neural network for generating small molecules similar to the ones in the training set. Our network consists of an encoder made up of bi-GRU layers for converting the input samples to a latent space, predictor for enhancing the capability of encoder made up of 1D-CNN layers and a decoder comprised of uni-GRU layers for reconstructing the samples from the latent space representation. Condition vector in latent space is used for generating molecules with the desired properties. We present the loss functions used for training our network, experimental details and property prediction metrics. Our network outperforms previous methods using Molecular weight, LogP and Quantitative Estimation of Drug-likeness as the evaluation metrics.
Electrode "pop" artifacts originate from the spontaneous loss of connectivity between a surface and an electrode. Electroencephalography (EEG) uses a dense array of electrodes, hence "popped" segments are among the most pervasive type of artifact seen during the collection of EEG data. In many cases, the continuity of EEG data is critical for downstream applications (e.g. brain machine interface) and requires that popped segments be accurately interpolated. In this paper we frame the interpolation problem as a self-learning task using a deep encoder-decoder network. We compare our approach against contemporary interpolation methods on a publicly available EEG data set. Our approach exhibited a minimum of ~15% improvement over contemporary approaches when tested on subjects and tasks not used during model training. We demonstrate how our model's performance can be enhanced further on novel subjects and tasks using transfer learning. All code and data associated with this study is open-source to enable ease of extension and practical use. To our knowledge, this work is the first solution to the EEG interpolation problem that uses deep learning.
Many top-performing image captioning models rely solely on object features computed with an object detection model to generate image descriptions. However, recent studies propose to directly use scene graphs to introduce information about object relations into captioning, hoping to better describe interactions between objects. In this work, we thoroughly investigate the use of scene graphs in image captioning. We empirically study whether using additional scene graph encoders can lead to better image descriptions and propose a conditional graph attention network (C-GAT), where the image captioning decoder state is used to condition the graph updates. Finally, we determine to what extent noise in the predicted scene graphs influence caption quality. Overall, we find no significant difference between models that use scene graph features and models that only use object detection features across different captioning metrics, which suggests that existing scene graph generation models are still too noisy to be useful in image captioning. Moreover, although the quality of predicted scene graphs is very low in general, when using high quality scene graphs we obtain gains of up to 3.3 CIDEr compared to a strong Bottom-Up Top-Down baseline. We open source code to reproduce all our experiments in https://github.com/iacercalixto/butd-image-captioning.
Automation of brain tumors in 3D magnetic resonance images (MRIs) is key to assess the diagnostic and treatment of the disease. In recent years, convolutional neural networks (CNNs) have shown improved results in the task. However, high memory consumption is still a problem in 3D-CNNs. Moreover, most methods do not include uncertainty information, which is specially critical in medical diagnosis. This work proposes a 3D encoder-decoder architecture, based on V-Net \cite{vnet} which is trained with patching techniques to reduce memory consumption and decrease the effect of unbalanced data. We also introduce voxel-wise uncertainty, both epistemic and aleatoric using test-time dropout and data-augmentation respectively. Uncertainty maps can provide extra information to expert neurologists, useful for detecting when the model is not confident on the provided segmentation.
To model diverse responses for a given post, one promising way is to introduce a latent variable into Seq2Seq models. The latent variable is supposed to capture the discourse-level information and encourage the informativeness of target responses. However, such discourse-level information is often too coarse for the decoder to be utilized. To tackle it, our idea is to transform the coarse-grained discourse-level information into fine-grained word-level information. Specifically, we firstly measure the semantic concentration of corresponding target response on the post words by introducing a fine-grained focus signal. Then, we propose a focus-constrained attention mechanism to take full advantage of focus in well aligning the input to the target response. The experimental results demonstrate that by exploiting the fine-grained signal, our model can generate more diverse and informative responses compared with several state-of-the-art models.
Retinal vessels are important biomarkers for many ophthalmological and cardiovascular diseases. Hence, it is of great significance to develop automatic models for computer-aided diagnosis. Existing methods, such as U-Net follow the encoder-decoder pipeline, where detailed information is lost in the encoder in order to achieve a large field of view. Although spatial detailed information could be recovered partly in the decoder, while there is noise in the high-resolution feature maps of the encoder. And, we argue this encoder-decoder architecture is inefficient for vessel segmentation. In this paper, we present the detail-preserving network (DPN), which avoids the encoder-decoder pipeline. To preserve detailed information and learn structural information simultaneously, we designed the detail-preserving block (DP-Block). Further, we stacked eight DP-Blocks together to form the DPN. More importantly, there are no down-sampling operations among these blocks. Therefore, the DPN could maintain a high/full resolution during processing, avoiding the loss of detailed information. To illustrate the effectiveness of DPN, we conducted experiments over three public datasets. Experimental results show, compared to state-of-the-art methods, DPN shows competitive/better performance in terms of segmentation accuracy, segmentation speed, and model size. Specifically, 1) Our method achieves comparable segmentation performance on the DRIVE, CHASE_DB1, and HRF datasets. 2) The segmentation speed of DPN is over 20-160 times faster than other methods on the DRIVE dataset. 3) The number of parameters of DPN is1 around 120k, far less than all comparison methods.
Current deep domain adaptation methods used in computer vision have mainly focused on learning discriminative and domain-invariant features across different domains. In this paper, we present a novel "deep adversarial transition learning" (DATL) framework that bridges the domain gap by projecting the source and target domains into intermediate, transitional spaces through the employment of adjustable, cross-grafted generative network stacks and effective adversarial learning between transitions. Specifically, we construct variational auto-encoders (VAE) for the two domains, and form bidirectional transitions by cross-grafting the VAEs' decoder stacks. Furthermore, generative adversarial networks (GAN) are employed for domain adaptation, mapping the target domain data to the known label space of the source domain. The overall adaptation process hence consists of three phases: feature representation learning by VAEs, transitions generation, and transitions alignment by GANs. Experimental results demonstrate that our method outperforms the state-of-the art on a number of unsupervised domain adaptation benchmarks.
Person re-identification (Re-ID) is a challenging task as persons are often in different backgrounds. Most recent Re-ID methods treat the foreground and background information equally for person discriminative learning, but can easily lead to potential false alarm problems when different persons are in similar backgrounds or the same person is in different backgrounds. In this paper, we propose a Foreground-Guided Texture-Focused Network (FTN) for Re-ID, which can weaken the representation of unrelated background and highlight the attributes person-related in an end-to-end manner. FTN consists of a semantic encoder (S-Enc) and a compact foreground attention module (CFA) for Re-ID task, and a texture-focused decoder (TF-Dec) for reconstruction task. Particularly, we build a foreground-guided semi-supervised learning strategy for TF-Dec because the reconstructed ground-truths are only the inputs of FTN weighted by the Gaussian mask and the attention mask generated by CFA. Moreover, a new gradient loss is introduced to encourage the network to mine the texture consistency between the inputs and the reconstructed outputs. Our FTN is computationally efficient and extensive experiments on three commonly used datasets Market1501, CUHK03 and MSMT17 demonstrate that the proposed method performs favorably against the state-of-the-art methods.
Video summarization is an effective way to facilitate video searching and browsing. Most of existing systems employ encoder-decoder based recurrent neural networks, which fail to explicitly diversify the system-generated summary frames while requiring intensive computations. In this paper, we propose an efficient convolutional neural network architecture for video SUMmarization via Global Diverse Attention called SUM-GDA, which adapts attention mechanism in a global perspective to consider pairwise temporal relations of video frames. Particularly, the GDA module has two advantages: 1) it models the relations within paired frames as well as the relations among all pairs, thus capturing the global attention across all frames of one video; 2) it reflects the importance of each frame to the whole video, leading to diverse attention on these frames. Thus, SUM-GDA is beneficial for generating diverse frames to form satisfactory video summary. Extensive experiments on three data sets, i.e., SumMe, TVSum, and VTW, have demonstrated that SUM-GDA and its extension outperform other competing state-of-the-art methods with remarkable improvements. In addition, the proposed models can be run in parallel with significantly less computational costs, which helps the deployment in highly demanding applications.
Recently, inspired by Transformer, self-attention-based scene text recognition approaches have achieved outstanding performance. However, we find that the size of model expands rapidly with the lexicon increasing. Specifically, the number of parameters for softmax classification layer and output embedding layer are proportional to the vocabulary size. It hinders the development of a lightweight text recognition model especially applied for Chinese and multiple languages. Thus, we propose a lightweight scene text recognition model named Hamming OCR. In this model, a novel Hamming classifier, which adopts locality sensitive hashing (LSH) algorithm to encode each character, is proposed to replace the softmax regression and the generated LSH code is directly employed to replace the output embedding. We also present a simplified transformer decoder to reduce the number of parameters by removing the feed-forward network and using cross-layer parameter sharing technique. Compared with traditional methods, the number of parameters in both classification and embedding layers is independent on the size of vocabulary, which significantly reduces the storage requirement without loss of accuracy. Experimental results on several datasets, including four public benchmaks and a Chinese text dataset synthesized by SynthText with more than 20,000 characters, shows that Hamming OCR achieves competitive results.
Computationally expensive temperature and power grid analyses are required during the design cycle to guide IC design. This paper employs encoder-decoder based generative (EDGe) networks to map these analyses to fast and accurate image-to-image and sequence-to-sequence translation tasks. The network takes a power map as input and outputs the corresponding temperature or IR drop map. We propose two networks: (i) ThermEDGe: a static and dynamic full-chip temperature estimator and (ii) IREDGe: a full-chip static IR drop predictor based on input power, power grid distribution, and power pad distribution patterns. The models are design-independent and must be trained just once for a particular technology and packaging solution. ThermEDGe and IREDGe are demonstrated to rapidly predict the on-chip temperature and IR drop contours in milliseconds (in contrast with commercial tools that require several hours or more) and provide an average error of 0.6% and 0.008% respectively.
This paper formulates the polar-code construction problem for the successive-cancellation list (SCL) decoder as a maze-traversing game, which can be solved by reinforcement learning techniques. The proposed method provides a novel technique for polar-code construction that no longer depends on sorting and selecting bit-channels by reliability. Instead, this technique decides whether the input bits should be frozen in a purely sequential manner. The equivalence of optimizing the polar-code construction for the SCL decoder under this technique and maximizing the expected reward of traversing a maze is drawn. Simulation results show that the standard polar-code constructions that are designed for the successive-cancellation decoder are no longer optimal for the SCL decoder with respect to the frame error rate. In contrast, the simulations show that, with a reasonable amount of training, the game-based construction method finds code constructions that have lower frame-error rate for various code lengths and decoders compared to standard constructions.
Commonsense knowledge graph (CKG) is a special type of knowledge graph (KG), where entities are composed of free-form text. However, most existing CKG completion methods focus on the setting where all the entities are presented at training time. Although this setting is standard for conventional KG completion, it has limitations for CKG completion. At test time, entities in CKGs can be unseen because they may have unseen text/names and entities may be disconnected from the training graph, since CKGs are generally very sparse. Here, we propose to study the inductive learning setting for CKG completion where unseen entities may present at test time. We develop a novel learning framework named InductivE. Different from previous approaches, InductiveE ensures the inductive learning capability by directly computing entity embeddings from raw entity attributes/text. InductiveE consists of a free-text encoder, a graph encoder, and a KG completion decoder. Specifically, the free-text encoder first extracts the textual representation of each entity based on the pre-trained language model and word embedding. The graph encoder is a gated relational graph convolutional neural network that learns from a densified graph for more informative entity representation learning. We develop a method that densifies CKGs by adding edges among semantic-related entities and provide more supportive information for unseen entities, leading to better generalization ability of entity embedding for unseen entities. Finally, inductiveE employs Conv-TransE as the CKG completion decoder. Experimental results show that InductiveE significantly outperforms state-of-the-art baselines in both standard and inductive settings on ATOMIC and ConceptNet benchmarks. InductivE performs especially well on inductive scenarios where it achieves above 48% improvement over present methods.
Dialogue policy learning for task-oriented dialogue systems has enjoyed great progress recently mostly through employing reinforcement learning methods. However, these approaches have become very sophisticated. It is time to re-evaluate it. Are we really making progress developing dialogue agents only based on reinforcement learning? We demonstrate how (1)~traditional supervised learning together with (2)~a simulator-free adversarial learning method can be used to achieve performance comparable to state-of-the-art RL-based methods. First, we introduce a simple dialogue action decoder to predict the appropriate actions. Then, the traditional multi-label classification solution for dialogue policy learning is extended by adding dense layers to improve the dialogue agent performance. Finally, we employ the Gumbel-Softmax estimator to alternatively train the dialogue agent and the dialogue reward model without using reinforcement learning. Based on our extensive experimentation, we can conclude the proposed methods can achieve more stable and higher performance with fewer efforts, such as the domain knowledge required to design a user simulator and the intractable parameter tuning in reinforcement learning. Our main goal is not to beat reinforcement learning with supervised learning, but to demonstrate the value of rethinking the role of reinforcement learning and supervised learning in optimizing task-oriented dialogue systems.
Neural Machine Translation (NMT) models often lack diversity in their generated translations, even when paired with search algorithm, like beam search. A challenge is that the diversity in translations are caused by the variability in the target language, and cannot be inferred from the source sentence alone. In this paper, we propose to explicitly model this one-to-many mapping by conditioning the decoder of a NMT model on a latent variable that represents the domain of target sentences. The domain is a discrete variable generated by a target encoder that is jointly trained with the NMT model. The predicted domain of target sentences are given as input to the decoder during training. At inference, we can generate diverse translations by decoding with different domains. Unlike our strongest baseline (Shen et al., 2019), our method can scale to any number of domains without affecting the performance or the training time. We assess the quality and diversity of translations generated by our model with several metrics, on three different datasets.
Representation of data on mixed variables, numerical and categorical types to get suitable feature map is a challenging task as important information lies in a complex non-linear manifold. The feature transformation should be able to incorporate marginal information of the individual variables and complex cross-dependence structure among the mixed type of variables simultaneously. In this work, we propose a novel nonlinear Deep Encoder-Decoder framework to capture the cross-domain information for mixed data types. The hidden layers of the network connect the two types of variables through various non-linear transformations to give latent feature maps. We encode the information on the numerical variables in a number of hidden nonlinear units. We use these units to recreate categorical variables through further nonlinear transformations. A separate and similar network is developed switching the roles of the numerical and categorical variables. The hidden representational units are stacked one next to the others and transformed into a common space using a locality preserving projection. The derived feature maps are used to explore the clusters in the data. Various standard datasets are investigated to show nearly the state of the art performance in clustering using the feature maps with simple K-means clustering.
Speech-to-text translation (ST), which directly translates the source language speech to the target language text, has attracted intensive attention recently. However, the combination of speech recognition and machine translation in a single model poses a heavy burden on the direct cross-modal cross-lingual mapping. To reduce the learning difficulty, we propose COnSecutive Transcription and Translation (COSTT), an integral approach for speech-to-text translation. The key idea is to generate source transcript and target translation text with a single decoder. It benefits the model training so that additional large parallel text corpus can be fully exploited to enhance the speech translation training. Our method is verified on three mainstream datasets, including Augmented LibriSpeech English-French dataset, IWSLT2018 English-German dataset, and TED English-Chinese dataset. Experiments show that our proposed COSTT outperforms or on par with the previous state-of-the-art methods on the three datasets. We have released our code at \url{https://github.com/dqqcasia/st}.