分类数据存在于健康或供应链等关键领域，此数据需要特定的治疗方法。为了将最新的机器学习模型应用于此类数据，需要编码。为了构建可解释的模型，单次编码仍然是一个很好的解决方案，但是这样的编码会创建稀疏的数据。梯度估计器不适合稀疏数据：梯度主要视为零，而它并不总是存在，因此引入了新型的梯度估计器。我们显示了该估计值在理论上最小化的内容，并在具有多个模型体系结构的不同数据集上显示了其效率。在相似的设置下，这种新的估计器的性能优于常见估计器。匿名后，现实世界零售数据集也会发布。总体而言，本文的目的是彻底考虑分类数据，并适应这些关键功能。

深度神经网络变得越来越强大，大大，并且始终需要培训更多标记的数据。但是，由于注释数据是耗时的，因此现在有必要开发在学习有限数据时显示出良好性能的系统。必须正确选择这些数据以获得仍然有效的模型。为此，系统必须能够确定应注释哪些数据以获得最佳结果。在本文中，我们提出了四个估计器来估计对象检测预测的信心。前两个基于蒙特卡洛辍学，第三个基于描述性统计，最后一个是检测器后验概率。在主动学习框架中，与随机选择图像相比，三个第一估计器在检测文档物理页面和文本线的性能方面有显着改善。我们还表明，基于描述性统计的提议估计器可以替代MC辍学，从而降低了计算成本而不会损害性能。

无约束的手写文本识别是一项具有挑战性的计算机视觉任务。传统上，它是通过两步方法来处理的，结合了线细分，然后是文本线识别。我们第一次为手写文档识别任务提出了无端到端的无分段体系结构：文档注意网络。除文本识别外，该模型还接受了使用类似XML的方式使用开始和结束标签标记文本零件的训练。该模型由用于特征提取的FCN编码器和用于复发令牌预测过程的变压器解码器层组成。它将整个文本文档作为输入和顺序输出字符以及逻辑布局令牌。与现有基于分割的方法相反，该模型是在不使用任何分割标签的情况下进行训练的。我们在页面级别的Read 2016数据集以及CER分别为3.43％和3.70％的双页级别上获得了竞争成果。我们还为Rimes 2009数据集提供了页面级别的结果，达到CER的4.54％。我们在https://github.com/factodeeplearning/dan上提供所有源代码和预训练的模型权重。

无约束的手写文本识别仍然具有挑战性的计算机视觉系统。段落识别传统上由两个模型实现：第一个用于线分割和用于文本线路识别的第二个。我们提出了一个统一的端到端模型，使用混合注意力来解决这项任务。该模型旨在迭代地通过线路进行段落图像线。它可以分为三个模块。编码器从整个段落图像生成特征映射。然后，注意力模块循环生成垂直加权掩模，使能专注于当前的文本线特征。这样，它执行一种隐式线分割。对于每个文本线特征，解码器模块识别关联的字符序列，导致整个段落的识别。我们在三个流行的数据集赛中达到最先进的字符错误率：ribs的1.91％，IAM 4.45％，读取2016年3.59％。我们的代码和培训的模型重量可在HTTPS：// GitHub上获得.com / fefodeeplearning / watermentattentocroc。

Convolutional Neural Networks (CNNs) have demonstrated superiority in learning patterns, but are sensitive to label noises and may overfit noisy labels during training. The early stopping strategy averts updating CNNs during the early training phase and is widely employed in the presence of noisy labels. Motivated by biological findings that the amplitude spectrum (AS) and phase spectrum (PS) in the frequency domain play different roles in the animal's vision system, we observe that PS, which captures more semantic information, can increase the robustness of DNNs to label noise, more so than AS can. We thus propose early stops at different times for AS and PS by disentangling the features of some layer(s) into AS and PS using Discrete Fourier Transform (DFT) during training. Our proposed Phase-AmplituDe DisentangLed Early Stopping (PADDLES) method is shown to be effective on both synthetic and real-world label-noise datasets. PADDLES outperforms other early stopping methods and obtains state-of-the-art performance.

We present a novel neural model for modern poetry generation in French. The model consists of two pretrained neural models that are fine-tuned for the poem generation task. The encoder of the model is a RoBERTa based one while the decoder is based on GPT-2. This way the model can benefit from the superior natural language understanding performance of RoBERTa and the good natural language generation performance of GPT-2. Our evaluation shows that the model can create French poetry successfully. On a 5 point scale, the lowest score of 3.57 was given by human judges to typicality and emotionality of the output poetry while the best score of 3.79 was given to understandability.

Graph Neural Networks (GNNs) have been successfully applied in many applications in computer sciences. Despite the success of deep learning architectures in other domains, deep GNNs still underperform their shallow counterparts. There are many open questions about deep GNNs, but over-smoothing and over-squashing are perhaps the most intriguing issues. When stacking multiple graph convolutional layers, the over-smoothing and over-squashing problems arise and have been defined as the inability of GNNs to learn deep representations and propagate information from distant nodes, respectively. Even though the widespread definitions of both problems are similar, these phenomena have been studied independently. This work strives to understand the underlying relationship between over-smoothing and over-squashing from a topological perspective. We show that both problems are intrinsically related to the spectral gap of the Laplacian of the graph. Therefore, there is a trade-off between these two problems, i.e., we cannot simultaneously alleviate both over-smoothing and over-squashing. We also propose a Stochastic Jost and Liu curvature Rewiring (SJLR) algorithm based on a bound of the Ollivier's Ricci curvature. SJLR is less expensive than previous curvature-based rewiring methods while retaining fundamental properties. Finally, we perform a thorough comparison of SJLR with previous techniques to alleviate over-smoothing or over-squashing, seeking to gain a better understanding of both problems.

We present a novel approach to generating news headlines in Finnish for a given news story. We model this as a summarization task where a model is given a news article, and its task is to produce a concise headline describing the main topic of the article. Because there are no openly available GPT-2 models for Finnish, we will first build such a model using several corpora. The model is then fine-tuned for the headline generation task using a massive news corpus. The system is evaluated by 3 expert journalists working in a Finnish media house. The results showcase the usability of the presented approach as a headline suggestion tool to facilitate the news production process.

We present a method for extracting a multilingual sentiment annotated dialog data set from Fallout New Vegas. The game developers have preannotated every line of dialog in the game in one of the 8 different sentiments: \textit{anger, disgust, fear, happy, neutral, pained, sad } and \textit{surprised}. The game has been translated into English, Spanish, German, French and Italian. We conduct experiments on multilingual, multilabel sentiment analysis on the extracted data set using multilingual BERT, XLMRoBERTa and language specific BERT models. In our experiments, multilingual BERT outperformed XLMRoBERTa for most of the languages, also language specific models were slightly better than multilingual BERT for most of the languages. The best overall accuracy was 54\% and it was achieved by using multilingual BERT on Spanish data. The extracted data set presents a challenging task for sentiment analysis. We have released the data, including the testing and training splits, openly on Zenodo. The data set has been shuffled for copyright reasons.

Word order, an essential property of natural languages, is injected in Transformer-based neural language models using position encoding. However, recent experiments have shown that explicit position encoding is not always useful, since some models without such feature managed to achieve state-of-the art performance on some tasks. To understand better this phenomenon, we examine the effect of removing position encodings on the pre-training objective itself (i.e., masked language modelling), to test whether models can reconstruct position information from co-occurrences alone. We do so by controlling the amount of masked tokens in the input sentence, as a proxy to affect the importance of position information for the task. We find that the necessity of position information increases with the amount of masking, and that masked language models without position encodings are not able to reconstruct this information on the task. These findings point towards a direct relationship between the amount of masking and the ability of Transformers to capture order-sensitive aspects of language using position encoding.