PyTorchKR Papers

최신 논문

총 9,642편
태그·기간 등 상세 필터

카테고리 탐색

로그인 후 이 태그 구독

arXiv:2109.08417

Transformer-Unet: Raw Image Processing with Unet

Medical image segmentation have drawn massive attention as it is important in biomedical image analysis. Good segmentation results can assist doctors with their judgement and further improve patients' experience. Among many available pipelines in medical image analysis, Unet is one of the most popular neural networks as it keeps raw features by adding concatenation between encoder and decoder, which makes it still widely used in industrial field. In the mean time, as a popular model which dominates natural language process tasks, transformer is now introduced to computer vision tasks and have seen promising results in object detection, image classification and semantic segmentation tasks. Therefore, the combination of transformer and Unet is supposed to be more efficient than both methods working individually. In this article, we propose Transformer-Unet by adding transformer modules in raw images instead of feature maps in Unet and test our network in CT82 datasets for Pancreas segmentation accordingly. We form an end-to-end network and gain segmentation results better than many previous Unet based algorithms in our experiment. We demonstrate our network and show our experimental results in this paper accordingly.

arXiv:2109.08247

Towards agricultural autonomy: crop row detection under varying field conditions using deep learning

This paper presents a novel metric to evaluate the robustness of deep learning based semantic segmentation approaches for crop row detection under different field conditions encountered by a field robot. A dataset with ten main categories encountered under various field conditions was used for testing. The effect on these conditions on the angular accuracy of crop row detection was compared. A deep convolutional encoder decoder network is implemented to predict crop row masks using RGB input images. The predicted mask is then sent to a post processing algorithm to extract the crop rows. The deep learning model was found to be robust against shadows and growth stages of the crop while the performance was reduced under direct sunlight, increasing weed density, tramlines and discontinuities in crop rows when evaluated with the novel metric.

arXiv:2109.08411

Cross Modification Attention Based Deliberation Model for Image Captioning

The conventional encoder-decoder framework for image captioning generally adopts a single-pass decoding process, which predicts the target descriptive sentence word by word in temporal order. Despite the great success of this framework, it still suffers from two serious disadvantages. Firstly, it is unable to correct the mistakes in the predicted words, which may mislead the subsequent prediction and result in error accumulation problem. Secondly, such a framework can only leverage the already generated words but not the possible future words, and thus lacks the ability of global planning on linguistic information. To overcome these limitations, we explore a universal two-pass decoding framework, where a single-pass decoding based model serving as the Drafting Model first generates a draft caption according to an input image, and a Deliberation Model then performs the polishing process to refine the draft caption to a better image description. Furthermore, inspired from the complementarity between different modalities, we propose a novel Cross Modification Attention (CMA) module to enhance the semantic expression of the image features and filter out error information from the draft captions. We integrate CMA with the decoder of our Deliberation Model and name it as Cross Modification Attention based Deliberation Model (CMA-DM). We train our proposed framework by jointly optimizing all trainable components from scratch with a trade-off coefficient. Experiments on MS COCO dataset demonstrate that our approach obtains significant improvements over single-pass decoding baselines and achieves competitive performances compared with other state-of-the-art two-pass decoding based methods.

arXiv:2109.08044

Eformer: Edge Enhancement based Transformer for Medical Image Denoising

In this work, we present Eformer - Edge enhancement based transformer, a novel architecture that builds an encoder-decoder network using transformer blocks for medical image denoising. Non-overlapping window-based self-attention is used in the transformer block that reduces computational requirements. This work further incorporates learnable Sobel-Feldman operators to enhance edges in the image and propose an effective way to concatenate them in the intermediate layers of our architecture. The experimental analysis is conducted by comparing deterministic learning and residual learning for the task of medical image denoising. To defend the effectiveness of our approach, our model is evaluated on the AAPM-Mayo Clinic Low-Dose CT Grand Challenge Dataset and achieves state-of-the-art performance, $i.e.$, 43.487 PSNR, 0.0067 RMSE, and 0.9861 SSIM. We believe that our work will encourage more research in transformer-based architectures for medical image denoising using residual learning.

arXiv:2109.07799공식 코드 · Papers with Code

Label-Attention Transformer with Geometrically Coherent Objects for Image Captioning

Automatic transcription of scene understanding in images and videos is a step towards artificial general intelligence. Image captioning is a nomenclature for describing meaningful information in an image using computer vision techniques. Automated image captioning techniques utilize encoder and decoder architecture, where the encoder extracts features from an image and the decoder generates a transcript. In this work, we investigate two unexplored ideas for image captioning using transformers: First, we demonstrate the enforcement of using objects' relevance in the surrounding environment. Second, learning an explicit association between labels and language constructs. We propose label-attention Transformer with geometrically coherent objects (LATGeO). The proposed technique acquires a proposal of geometrically coherent objects using a deep neural network (DNN) and generates captions by investigating their relationships using a label-attention module. Object coherence is defined using the localized ratio of the geometrical properties of the proposals. The label-attention module associates the extracted objects classes to the available dictionary using self-attention layers. The experimentation results show that objects' relevance in surroundings and binding of their visual feature with their geometrically localized ratios combined with its associated labels help in defining meaningful captions. The proposed framework is tested on the MSCOCO dataset, and a thorough evaluation resulting in overall better quantitative scores pronounces its superiority.

arXiv:2109.08008공식 코드 · Papers with Code

The NiuTrans System for WNGT 2020 Efficiency Task

This paper describes the submissions of the NiuTrans Team to the WNGT 2020 Efficiency Shared Task. We focus on the efficient implementation of deep Transformer models \cite{wang-etal-2019-learning, li-etal-2019-niutrans} using NiuTensor (https://github.com/NiuTrans/NiuTensor), a flexible toolkit for NLP tasks. We explored the combination of deep encoder and shallow decoder in Transformer models via model compression and knowledge distillation. The neural machine translation decoding also benefits from FP16 inference, attention caching, dynamic batching, and batch pruning. Our systems achieve promising results in both translation quality and efficiency, e.g., our fastest system can translate more than 40,000 tokens per second with an RTX 2080 Ti while maintaining 42.9 BLEU on \textit{newstest2018}. The code, models, and docker images are available at NiuTrans.NMT (https://github.com/NiuTrans/NiuTrans.NMT).

arXiv:2109.07812공식 코드 · Papers with Code

Transductive Learning for Unsupervised Text Style Transfer

Unsupervised style transfer models are mainly based on an inductive learning approach, which represents the style as embeddings, decoder parameters, or discriminator parameters and directly applies these general rules to the test cases. However, the lacking of parallel corpus hinders the ability of these inductive learning methods on this task. As a result, it is likely to cause severe inconsistent style expressions, like `the salad is rude`. To tackle this problem, we propose a novel transductive learning approach in this paper, based on a retrieval-based context-aware style representation. Specifically, an attentional encoder-decoder with a retriever framework is utilized. It involves top-K relevant sentences in the target style in the transfer process. In this way, we can learn a context-aware style embedding to alleviate the above inconsistency problem. In this paper, both sparse (BM25) and dense retrieval functions (MIPS) are used, and two objective functions are designed to facilitate joint learning. Experimental results show that our method outperforms several strong baselines. The proposed transductive learning approach is general and effective to the task of unsupervised style transfer, and we will apply it to the other two typical methods in the future.

arXiv:2109.07740

Scaling Laws for Neural Machine Translation

We present an empirical study of scaling properties of encoder-decoder Transformer models used in neural machine translation (NMT). We show that cross-entropy loss as a function of model size follows a certain scaling law. Specifically (i) We propose a formula which describes the scaling behavior of cross-entropy loss as a bivariate function of encoder and decoder size, and show that it gives accurate predictions under a variety of scaling approaches and languages; we show that the total number of parameters alone is not sufficient for such purposes. (ii) We observe different power law exponents when scaling the decoder vs scaling the encoder, and provide recommendations for optimal allocation of encoder/decoder capacity based on this observation. (iii) We also report that the scaling behavior of the model is acutely influenced by composition bias of the train/test sets, which we define as any deviation from naturally generated text (either via machine generated or human translated text). We observe that natural text on the target side enjoys scaling, which manifests as successful reduction of the cross-entropy loss. (iv) Finally, we investigate the relationship between the cross-entropy loss and the quality of the generated translations. We find two different behaviors, depending on the nature of the test data. For test sets which were originally translated from target language to source language, both loss and BLEU score improve as model size increases. In contrast, for test sets originally translated from source language to target language, the loss improves, but the BLEU score stops improving after a certain threshold. We release generated text from all models used in this study.

arXiv:2109.07513

Tied & Reduced RNN-T Decoder

Previous works on the Recurrent Neural Network-Transducer (RNN-T) models have shown that, under some conditions, it is possible to simplify its prediction network with little or no loss in recognition accuracy (arXiv:2003.07705 [eess.AS], [2], arXiv:2012.06749 [cs.CL]). This is done by limiting the context size of previous labels and/or using a simpler architecture for its layers instead of LSTMs. The benefits of such changes include reduction in model size, faster inference and power savings, which are all useful for on-device applications. In this work, we study ways to make the RNN-T decoder (prediction network + joint network) smaller and faster without degradation in recognition performance. Our prediction network performs a simple weighted averaging of the input embeddings, and shares its embedding matrix weights with the joint network's output layer (a.k.a. weight tying, commonly used in language modeling arXiv:1611.01462 [cs.LG]). This simple design, when used in conjunction with additional Edit-based Minimum Bayes Risk (EMBR) training, reduces the RNN-T Decoder from 23M parameters to just 2M, without affecting word-error rate (WER).

arXiv:2109.07368공식 코드 · Papers with Code

Learning When to Translate for Streaming Speech

How to find proper moments to generate partial sentence translation given a streaming speech input? Existing approaches waiting-and-translating for a fixed duration often break the acoustic units in speech, since the boundaries between acoustic units in speech are not even. In this paper, we propose MoSST, a simple yet effective method for translating streaming speech content. Given a usually long speech sequence, we develop an efficient monotonic segmentation module inside an encoder-decoder model to accumulate acoustic information incrementally and detect proper speech unit boundaries for the input in speech translation task. Experiments on multiple translation directions of the MuST-C dataset show that MoSST outperforms existing methods and achieves the best trade-off between translation quality (BLEU) and latency. Our code is available at https://github.com/dqqcasia/mosst.

arXiv:2109.07043공식 코드 · Papers with Code

Attention Is Indeed All You Need: Semantically Attention-Guided Decoding for Data-to-Text NLG

Ever since neural models were adopted in data-to-text language generation, they have invariably been reliant on extrinsic components to improve their semantic accuracy, because the models normally do not exhibit the ability to generate text that reliably mentions all of the information provided in the input. In this paper, we propose a novel decoding method that extracts interpretable information from encoder-decoder models' cross-attention, and uses it to infer which attributes are mentioned in the generated text, which is subsequently used to rescore beam hypotheses. Using this decoding method with T5 and BART, we show on three datasets its ability to dramatically reduce semantic errors in the generated outputs, while maintaining their state-of-the-art quality.

arXiv:2109.07006

A Three Step Training Approach with Data Augmentation for Morphological Inflection

We present the BME submission for the SIGMORPHON 2021 Task 0 Part 1, Generalization Across Typologically Diverse Languages shared task. We use an LSTM encoder-decoder model with three step training that is first trained on all languages, then fine-tuned on each language families and finally finetuned on individual languages. We use a different type of data augmentation technique in the first two steps. Our system outperformed the only other submission. Although it remains worse than the Transformer baseline released by the organizers, our model is simpler and our data augmentation techniques are easily applicable to new languages. We perform ablation studies and show that the augmentation techniques and the three training steps often help but sometimes have a negative effect.

arXiv:2109.07295

New Perspective on Progressive GANs Distillation for One-class Novelty Detection

One-class novelty detection is conducted to identify anomalous instances, with different distributions from the expected normal instances. In this paper, the Generative Adversarial Network based on the Encoder-Decoder-Encoder scheme (EDE-GAN) achieves state-of-the-art performance. The two factors bellow serve the above purpose: 1) The EDE-GAN calculates the distance between two latent vectors as the anomaly score, which is unlike the previous methods by utilizing the reconstruction error between images. 2) The model obtains best results when the batch size is set to 1. To illustrate their superiority, we design a new GAN architecture, and compare performances according to different batch sizes. Moreover, with experimentation leads to discovery, our result implies there is also evidence of just how beneficial constraint on the latent space are when engaging in model training. In an attempt to learn compact and fast models, we present a new technology, Progressive Knowledge Distillation with GANs (P-KDGAN), which connects two standard GANs through the designed distillation loss. Two-step progressive learning continuously augments the performance of student GANs with improved results over single-step approach. Our experimental results on CIFAR-10, MNIST, and FMNIST datasets illustrate that P-KDGAN improves the performance of the student GAN by 2.44%, 1.77%, and 1.73% when compressing the computationat ratios of 24.45:1, 311.11:1, and 700:1, respectively.

arXiv:2109.07045

Uncertainty Quantification in Medical Image Segmentation with Multi-decoder U-Net

Accurate medical image segmentation is crucial for diagnosis and analysis. However, the models without calibrated uncertainty estimates might lead to errors in downstream analysis and exhibit low levels of robustness. Estimating the uncertainty in the measurement is vital to making definite, informed conclusions. Especially, it is difficult to make accurate predictions on ambiguous areas and focus boundaries for both models and radiologists, even harder to reach a consensus with multiple annotations. In this work, the uncertainty under these areas is studied, which introduces significant information with anatomical structure and is as important as segmentation performance. We exploit the medical image segmentation uncertainty quantification by measuring segmentation performance with multiple annotations in a supervised learning manner and propose a U-Net based architecture with multiple decoders, where the image representation is encoded with the same encoder, and segmentation referring to each annotation is estimated with multiple decoders. Nevertheless, a cross-loss function is proposed for bridging the gap between different branches. The proposed architecture is trained in an end-to-end manner and able to improve predictive uncertainty estimates. The model achieves comparable performance with fewer parameters to the integrated training model that ranked the runner-up in the MICCAI-QUBIQ 2020 challenge.

arXiv:2109.06717

Controllable Dialogue Generation with Disentangled Multi-grained Style Specification and Attribute Consistency Reward

Controllable text generation is an appealing but challenging task, which allows users to specify particular attributes of the generated outputs. In this paper, we propose a controllable dialogue generation model to steer response generation under multi-attribute constraints. Specifically, we define and categorize the commonly used control attributes into global and local ones, which possess different granularities of effects on response generation. Then, we significantly extend the conventional seq2seq framework by introducing a novel two-stage decoder, which first uses a multi-grained style specification layer to impose the stylistic constraints and determine word-level control states of responses based on the attributes, and then employs a response generation layer to generate final responses maintaining both semantic relevancy to the contexts and fidelity to the attributes. Furthermore, we train our model with an attribute consistency reward to promote response control with explicit supervision signals. Extensive experiments and in-depth analyses on two datasets indicate that our model can significantly outperform competitive baselines in terms of response quality, content diversity and controllability.

arXiv:2109.06704

KFCNet: Knowledge Filtering and Contrastive Learning Network for Generative Commonsense Reasoning

Pre-trained language models have led to substantial gains over a broad range of natural language processing (NLP) tasks, but have been shown to have limitations for natural language generation tasks with high-quality requirements on the output, such as commonsense generation and ad keyword generation. In this work, we present a novel Knowledge Filtering and Contrastive learning Network (KFCNet) which references external knowledge and achieves better generation performance. Specifically, we propose a BERT-based filter model to remove low-quality candidates, and apply contrastive learning separately to each of the encoder and decoder, within a general encoder--decoder architecture. The encoder contrastive module helps to capture global target semantics during encoding, and the decoder contrastive module enhances the utility of retrieved prototypes while learning general features. Extensive experiments on the CommonGen benchmark show that our model outperforms the previous state of the art by a large margin: +6.6 points (42.5 vs. 35.9) for BLEU-4, +3.7 points (33.3 vs. 29.6) for SPICE, and +1.3 points (18.3 vs. 17.0) for CIDEr. We further verify the effectiveness of the proposed contrastive module on ad keyword generation, and show that our model has potential commercial value.

arXiv:2109.06684

Non-autoregressive Transformer with Unified Bidirectional Decoder for Automatic Speech Recognition

Non-autoregressive (NAR) transformer models have been studied intensively in automatic speech recognition (ASR), and a substantial part of NAR transformer models is to use the casual mask to limit token dependencies. However, the casual mask is designed for the left-to-right decoding process of the non-parallel autoregressive (AR) transformer, which is inappropriate for the parallel NAR transformer since it ignores the right-to-left contexts. Some models are proposed to utilize right-to-left contexts with an extra decoder, but these methods increase the model complexity. To tackle the above problems, we propose a new non-autoregressive transformer with a unified bidirectional decoder (NAT-UBD), which can simultaneously utilize left-to-right and right-to-left contexts. However, direct use of bidirectional contexts will cause information leakage, which means the decoder output can be affected by the character information from the input of the same position. To avoid information leakage, we propose a novel attention mask and modify vanilla queries, keys, and values matrices for NAT-UBD. Experimental results verify that NAT-UBD can achieve character error rates (CERs) of 5.0%/5.5% on the Aishell1 dev/test sets, outperforming all previous NAR transformer models. Moreover, NAT-UBD can run 49.8x faster than the AR transformer baseline when decoding in a single step.

arXiv:2109.06679

Efficient Inference for Multilingual Neural Machine Translation

Multilingual NMT has become an attractive solution for MT deployment in production. But to match bilingual quality, it comes at the cost of larger and slower models. In this work, we consider several ways to make multilingual NMT faster at inference without degrading its quality. We experiment with several "light decoder" architectures in two 20-language multi-parallel settings: small-scale on TED Talks and large-scale on ParaCrawl. Our experiments demonstrate that combining a shallow decoder with vocabulary filtering leads to more than twice faster inference with no loss in translation quality. We validate our findings with BLEU and chrF (on 380 language pairs), robustness evaluation and human evaluation.

arXiv:2109.06481

AligNART: Non-autoregressive Neural Machine Translation by Jointly Learning to Estimate Alignment and Translate

Non-autoregressive neural machine translation (NART) models suffer from the multi-modality problem which causes translation inconsistency such as token repetition. Most recent approaches have attempted to solve this problem by implicitly modeling dependencies between outputs. In this paper, we introduce AligNART, which leverages full alignment information to explicitly reduce the modality of the target distribution. AligNART divides the machine translation task into $(i)$ alignment estimation and $(ii)$ translation with aligned decoder inputs, guiding the decoder to focus on simplified one-to-one translation. To alleviate the alignment estimation problem, we further propose a novel alignment decomposition method. Our experiments show that AligNART outperforms previous non-iterative NART models that focus on explicit modality reduction on WMT14 En$\leftrightarrow$De and WMT16 Ro$\rightarrow$En. Furthermore, AligNART achieves BLEU scores comparable to those of the state-of-the-art connectionist temporal classification based models on WMT14 En$\leftrightarrow$De. We also observe that AligNART effectively addresses the token repetition problem even without sequence-level knowledge distillation.

arXiv:2109.06388

On Distributed Learning with Constant Communication Bits

In this paper, we study a distributed learning problem constrained by constant communication bits. Specifically, we consider the distributed hypothesis testing (DHT) problem where two distributed nodes are constrained to transmit a constant number of bits to a central decoder. In such cases, we show that in order to achieve the optimal error exponents, it suffices to consider the empirical distributions of observed data sequences and encode them to the transmission bits. With such a coding strategy, we develop a geometric approach in the distribution spaces and establish an inner bound of error exponent regions. In particular, we show the optimal achievable error exponents and coding schemes for the following cases: (i) both nodes can transmit $\log_23$ bits; (ii) one of the nodes can transmit $1$ bit, and the other node is not constrained; (iii) the joint distribution of the nodes are conditionally independent given one hypothesis. Furthermore, we provide several numerical examples for illustrating the theoretical results. Our results provide theoretical guidance for designing practical distributed learning rules, and the developed approach also reveals new potentials for establishing error exponents for DHT with more general communication constraints.