Non-Intrusive Load Monitoring (NILM) aims to predict the status or consumption of domestic appliances in a household only by knowing the aggregated power load. NILM can be formulated as regression problem or most often as a classification problem. Most datasets gathered by smart meters allow to define naturally a regression problem, but the corresponding classification problem is a derived one, since it requires a conversion from the power signal to the status of each device by a thresholding method. We treat three different thresholding methods to perform this task, discussing their differences on various devices from the UK-DALE dataset. We analyze the performance of deep learning state-of-the-art architectures on both the regression and classification problems, introducing criteria to select the most convenient thresholding method.
Graph Neural Networks (GNNs) are effective in many applications. Still, there is a limited understanding of the effect of common graph structures on the learning process of GNNs. In this work, we systematically study the impact of community structure on the performance of GNNs in semi-supervised node classification on graphs. Following an ablation study on six datasets, we measure the performance of GNNs on the original graphs, and the change in performance in the presence and the absence of community structure. Our results suggest that communities typically have a major impact on the learning process and classification performance. For example, in cases where the majority of nodes from one community share a single classification label, breaking up community structure results in a significant performance drop. On the other hand, for cases where labels show low correlation with communities, we find that the graph structure is rather irrelevant to the learning process, and a feature-only baseline becomes hard to beat. With our work, we provide deeper insights in the abilities and limitations of GNNs, including a set of general guidelines for model selection based on the graph structure.
Natural language data exhibit tree-like hierarchical structures such as the hypernym-hyponym relations in WordNet. FastText, as the state-of-the-art text classifier based on shallow neural network in Euclidean space, may not model such hierarchies precisely with limited representation capacity. Considering that hyperbolic space is naturally suitable for modeling tree-like hierarchical data, we propose a new model named HyperText for efficient text classification by endowing FastText with hyperbolic geometry. Empirically, we show that HyperText outperforms FastText on a range of text classification tasks with much reduced parameters.
Due to increasing threats from malicious software (malware) in both number and complexity, researchers have developed approaches to automatic detection and classification of malware, instead of analyzing methods for malware files manually in a time-consuming effort. At the same time, malware authors have developed techniques to evade signature-based detection techniques used by antivirus companies. Most recently, deep learning is being used in malware classification to solve this issue. In this paper, we use several convolutional neural network (CNN) models for static malware classification. In particular, we use six deep learning models, three of which are past winners of the ImageNet Large-Scale Visual Recognition Challenge. The other three models are CNN-SVM, GRU-SVM and MLP-SVM, which enhance neural models with support vector machines (SVM). We perform experiments using the Malimg dataset, which has malware images that were converted from Portable Executable malware binaries. The dataset is divided into 25 malware families. Comparisons show that the Inception V3 model achieves a test accuracy of 99.24%, which is better than the accuracy of 98.52% achieved by the current state-of-the-art system called the M-CNN model.
Contrastive learning (CL) has been successful as a powerful representation learning method. In this work we propose CLIM: Contrastive Learning with mutual Information Maximization, to explore the potential of CL on cross-domain sentiment classification. To the best of our knowledge, CLIM is the first to adopt contrastive learning for natural language processing (NLP) tasks across domains. Due to scarcity of labels on the target domain, we introduce mutual information maximization (MIM) apart from CL to exploit the features that best support the final prediction. Furthermore, MIM is able to maintain a relatively balanced distribution of the model's prediction, and enlarges the margin between classes on the target domain. The larger margin increases our model's robustness and enables the same classifier to be optimal across domains. Consequently, we achieve new state-of-the-art results on the Amazon-review dataset as well as the airlines dataset, showing the efficacy of our proposed method CLIM.
Relation classification aims to extract semantic relations between entity pairs from the sentences. However, most existing methods can only identify seen relation classes that occurred during training. To recognize unseen relations at test time, we explore the problem of zero-shot relation classification. Previous work regards the problem as reading comprehension or textual entailment, which have to rely on artificial descriptive information to improve the understandability of relation types. Thus, rich semantic knowledge of the relation labels is ignored. In this paper, we propose a novel logic-guided semantic representation learning model for zero-shot relation classification. Our approach builds connections between seen and unseen relations via implicit and explicit semantic representations with knowledge graph embeddings and logic rules. Extensive experimental results demonstrate that our method can generalize to unseen relation types and achieve promising improvements.
Motor-Imagery based BCI (MI-BCI) neurorehabilitation can improve locomotor ability and reduce the deficit symptoms in Parkinson's Disease patients. Advanced Motor-Imagery BCI methods are needed to overcome the accuracy and time-related MI BCI calibration challenges in such patients. In this study, we proposed a Multi-session FBCSP (msFBCSP) based on inter-session transfer learning and we investigated its performance compared to the single-session based FBSCP. The main result of this study is the significantly improved accuracy obtained by proposed msFBCSP compared to single-session FBCSP in PD patients (median 81.3%, range 41.2-100.0% vs median 61.1%, range 25.0-100.0%, respectively; p<0.001). In conclusion, this study proposes a transfer learning-based multi-session based FBCSP approach which allowed to significantly improve calibration accuracy in MI BCI performed on PD patients.
Previous work has proposed many new loss functions and regularizers that improve test accuracy on image classification tasks. However, it is not clear whether these loss functions learn better representations for downstream tasks. This paper studies how the choice of training objective affects the transferability of the hidden representations of convolutional neural networks trained on ImageNet. We show that many objectives lead to statistically significant improvements in ImageNet accuracy over vanilla softmax cross-entropy, but the resulting fixed feature extractors transfer substantially worse to downstream tasks, and the choice of loss has little effect when networks are fully fine-tuned on the new tasks. Using centered kernel alignment to measure similarity between hidden representations of networks, we find that differences among loss functions are apparent only in the last few layers of the network. We delve deeper into representations of the penultimate layer, finding that different objectives and hyperparameter combinations lead to dramatically different levels of class separation. Representations with higher class separation obtain higher accuracy on the original task, but their features are less useful for downstream tasks. Our results suggest there exists a trade-off between learning invariant features for the original task and features relevant for transfer tasks.
Renal cell carcinoma (RCC) is the most common renal cancer in adults. The histopathologic classification of RCC is essential for diagnosis, prognosis, and management of patients. Reorganization and classification of complex histologic patterns of RCC on biopsy and surgical resection slides under a microscope remains a heavily specialized, error-prone, and time-consuming task for pathologists. In this study, we developed a deep neural network model that can accurately classify digitized surgical resection slides and biopsy slides into five related classes: clear cell RCC, papillary RCC, chromophobe RCC, renal oncocytoma, and normal. In addition to the whole-slide classification pipeline, we visualized the identified indicative regions and features on slides for classification by reprocessing patch-level classification results to ensure the explainability of our diagnostic model. We evaluated our model on independent test sets of 78 surgical resection whole slides and 79 biopsy slides from our tertiary medical institution, and 69 randomly selected surgical resection slides from The Cancer Genome Atlas (TCGA) database. The average area under the curve (AUC) of our classifier on the internal resection slides, internal biopsy slides, and external TCGA slides is 0.98, 0.98 and 0.99, respectively. Our results suggest that the high generalizability of our approach across different data sources and specimen types. More importantly, our model has the potential to assist pathologists by (1) automatically pre-screening slides to reduce false-negative cases, (2) highlighting regions of importance on digitized slides to accelerate diagnosis, and (3) providing objective and accurate diagnosis as the second opinion.
In this paper, we provide a theory of using graph neural networks (GNNs) for multi-node representation learning (where we are interested in learning a representation for a set of more than one node, such as link). We know that GNN is designed to learn single-node representations. When we want to learn a node set representation involving multiple nodes, a common practice in previous works is to directly aggregate the single-node representations obtained by a GNN into a joint node set representation. In this paper, we show a fundamental constraint of such an approach, namely the inability to capture the dependence between nodes in the node set, and argue that directly aggregating individual node representations does not lead to an effective joint representation for multiple nodes. Then, we notice that a few previous successful works for multi-node representation learning, including SEAL, Distance Encoding, and ID-GNN, all used node labeling. These methods first label nodes in the graph according to their relationships with the target node set before applying a GNN. Then, the node representations obtained in the labeled graph are aggregated into a node set representation. By investigating their inner mechanisms, we unify these node labeling techniques into a single and most general form -- labeling trick. We prove that with labeling trick a sufficiently expressive GNN learns the most expressive node set representations, thus in principle solves any joint learning tasks over node sets. Experiments on one important two-node representation learning task, link prediction, verified our theory. Our work explains the superior performance of previous node-labeling-based methods, and establishes a theoretical foundation of using GNNs for multi-node representation learning.
One of the main problems with biomedical signals is the limited amount of patient-specific data and the significant amount of time needed to record the sufficient number of samples needed for diagnostic and treatment purposes. In this study, we present a framework to simultaneously generate and classify biomedical time series based on a modified Adversarial Autoencoder (AAE) algorithm and one-dimensional convolutions. Our work is based on breathing time series, with specific motivation to capture breathing motion during radiotherapy lung cancer treatments. First, we explore the potential in using the Variational Autoencoder (VAE) and AAE algorithms to model breathing from individual patients. We extend the AAE algorithm to allow joint semi-supervised classification and generation of different types of signals. To simplify the modeling task, we introduce a pre-processing and post-processing compressing algorithm that transforms the multi-dimensional time series into vectors containing time and position values, which are transformed back into time series through an additional neural network. By incorporating few labeled samples during training, our model outperforms other purely discriminative networks in classifying breathing baseline shift irregularities from a dataset completely different from the training set. To our knowledge, the presented framework is the first approach that unifies generation and classification within a single model for this type of biomedical data, enabling both computer aided diagnosis and augmentation of labeled samples within a single framework.
Artificial intelligent (AI) algorithms, such as deep learning and XGboost, are used in numerous applications including computer vision, autonomous driving, and medical diagnostics. The robustness of these AI algorithms is of great interest as inaccurate prediction could result in safety concerns and limit the adoption of AI systems. In this paper, we propose a framework based on design of experiments to systematically investigate the robustness of AI classification algorithms. A robust classification algorithm is expected to have high accuracy and low variability under different application scenarios. The robustness can be affected by a wide range of factors such as the imbalance of class labels in the training dataset, the chosen prediction algorithm, the chosen dataset of the application, and a change of distribution in the training and test datasets. To investigate the robustness of AI classification algorithms, we conduct a comprehensive set of mixture experiments to collect prediction performance results. Then statistical analyses are conducted to understand how various factors affect the robustness of AI classification algorithms. We summarize our findings and provide suggestions to practitioners in AI applications.
The 12-lead electrocardiogram (ECG) is a commonly used tool for detecting cardiac abnormalities such as atrial fibrillation, blocks, and irregular complexes. For the PhysioNet/CinC 2020 Challenge, we built an algorithm using gradient boosted tree ensembles fitted on morphology and signal processing features to classify ECG diagnosis. For each lead, we derive features from heart rate variability, PQRST template shape, and the full signal waveform. We join the features of all 12 leads to fit an ensemble of gradient boosting decision trees to predict probabilities of ECG instances belonging to each class. We train a phase one set of feature importance determining models to isolate the top 1,000 most important features to use in our phase two diagnosis prediction models. We use repeated random sub-sampling by splitting our dataset of 43,101 records into 100 independent runs of 85:15 training/validation splits for our internal evaluation results. Our methodology generates us an official phase validation set score of 0.476 and test set score of -0.080 under the team name, CVC, placing us 36 out of 41 in the rankings.
Zero-shot classification is a generalization task where no instance from the target classes is seen during training. To allow for test-time transfer, each class is annotated with semantic information, commonly in the form of attributes or text descriptions. While classical zero-shot learning does not explicitly forbid using information from other datasets, the approaches that achieve the best absolute performance on image benchmarks rely on features extracted from encoders pretrained on Imagenet. This approach relies on hyper-optimized Imagenet-relevant parameters from the supervised classification setting, entangling important questions about the suitability of those parameters and how they were learned with more fundamental questions about representation learning and generalization. To remove these distractors, we propose a more challenging setting: Zero-Shot Learning from scratch (ZFS), which explicitly forbids the use of encoders fine-tuned on other datasets. Our analysis on this setting highlights the importance of local information, and compositional representations.
Deep learning has shown outstanding performance in several applications including image classification. However, deep classifiers are known to be highly vulnerable to adversarial attacks, in that a minor perturbation of the input can easily lead to an error. Providing robustness to adversarial attacks is a very challenging task especially in problems involving a large number of classes, as it typically comes at the expense of an accuracy decrease. In this work, we propose the Gaussian class-conditional simplex (GCCS) loss: a novel approach for training deep robust multiclass classifiers that provides adversarial robustness while at the same time achieving or even surpassing the classification accuracy of state-of-the-art methods. Differently from other frameworks, the proposed method learns a mapping of the input classes onto target distributions in a latent space such that the classes are linearly separable. Instead of maximizing the likelihood of target labels for individual samples, our objective function pushes the network to produce feature distributions yielding high inter-class separation. The mean values of the distributions are centered on the vertices of a simplex such that each class is at the same distance from every other class. We show that the regularization of the latent space based on our approach yields excellent classification accuracy and inherently provides robustness to multiple adversarial attacks, both targeted and untargeted, outperforming state-of-the-art approaches over challenging datasets.
Lung cancer classification in screening computed tomography (CT) scans is one of the most crucial tasks for early detection of this disease. Many lives can be saved if we are able to accurately classify malignant/cancerous lung nodules. Consequently, several deep learning based models have been proposed recently to classify lung nodules as malignant or benign. Nevertheless, the large variation in the size and heterogeneous appearance of the nodules makes this task an extremely challenging one. We propose a new Progressive Growing Channel Attentive Non-Local (ProCAN) network for lung nodule classification. The proposed method addresses this challenge from three different aspects. First, we enrich the Non-Local network by adding channel-wise attention capability to it. Second, we apply Curriculum Learning principles, whereby we first train our model on easy examples before hard ones. Third, as the classification task gets harder during the Curriculum learning, our model is progressively grown to increase its capability of handling the task at hand. We examined our proposed method on two different public datasets and compared its performance with state-of-the-art methods in the literature. The results show that the ProCAN model outperforms state-of-the-art methods and achieves an AUC of 98.05% and an accuracy of 95.28% on the LIDC-IDRI dataset. Moreover, we conducted extensive ablation studies to analyze the contribution and effects of each new component of our proposed method.
Despite the enormous success of machine learning models in various applications, most of these models lack resilience to (even small) perturbations in their input data. Hence, new methods to robustify machine learning models seem very essential. To this end, in this paper we consider the problem of binary classification with adversarial perturbations. Investigating the solution to a min-max optimization (which considers the worst-case loss in the presence of adversarial perturbations) we introduce a generalization to the max-margin classifier which takes into account the power of the adversary in manipulating the data. We refer to this classifier as the "Robust Max-margin" (RM) classifier. Under some mild assumptions on the loss function, we theoretically show that the gradient descent iterates (with sufficiently small step size) converge to the RM classifier in its direction. Therefore, the RM classifier can be studied to compute various performance measures (e.g. generalization error) of binary classification with adversarial perturbations.
Although well-trained deep neural networks have shown remarkable performance on numerous tasks, they rapidly forget what they have learned as soon as they begin to learn with additional data with the previous data stop being provided. In this paper, we introduce a novel algorithm, Incremental Class Learning with Attribute Sharing (ICLAS), for incremental class learning with deep neural networks. As one of its component, we also introduce a generative model, incGAN, which can generate images with increased variety compared with the training data. Under challenging environment of data deficiency, ICLAS incrementally trains classification and the generation networks. Since ICLAS trains both networks, our algorithm can perform multiple times of incremental class learning. The experiments on MNIST dataset demonstrate the advantages of our algorithm.
Multiple Sclerosis (MS) is a chronic, inflammatory and degenerative neurological disease, which is monitored by a specialist using the Expanded Disability Status Scale (EDSS) and recorded in unstructured text in the form of a neurology consult note. An EDSS measurement contains an overall "EDSS" score and several functional subscores. Typically, expert knowledge is required to interpret consult notes and generate these scores. Previous approaches used limited context length Word2Vec embeddings and keyword searches to predict scores given a consult note, but often failed when scores were not explicitly stated. In this work, we present MS-BERT, the first publicly available transformer model trained on real clinical data other than MIMIC. Next, we present MSBC, a classifier that applies MS-BERT to generate embeddings and predict EDSS and functional subscores. Lastly, we explore combining MSBC with other models through the use of Snorkel to generate scores for unlabelled consult notes. MSBC achieves state-of-the-art performance on all metrics and prediction tasks and outperforms the models generated from the Snorkel ensemble. We improve Macro-F1 by 0.12 (to 0.88) for predicting EDSS and on average by 0.29 (to 0.63) for predicting functional subscores over previous Word2Vec CNN and rule-based approaches.
Semantically-aligned $(speech, image)$ datasets can be used to explore "visually-grounded speech". In a majority of existing investigations, features of an image signal are extracted using neural networks "pre-trained" on other tasks (e.g., classification on ImageNet). In still others, pre-trained networks are used to extract audio features prior to semantic embedding. Without "transfer learning" through pre-trained initialization or pre-trained feature extraction, previous results have tended to show low rates of recall in $speech \rightarrow image$ and $image \rightarrow speech$ queries. Choosing appropriate neural architectures for encoders in the speech and image branches and using large datasets, one can obtain competitive recall rates without any reliance on any pre-trained initialization or feature extraction: $(speech,image)$ semantic alignment and $speech \rightarrow image$ and $image \rightarrow speech$ retrieval are canonical tasks worthy of independent investigation of their own and allow one to explore other questions---e.g., the size of the audio embedder can be reduced significantly with little loss of recall rates in $speech \rightarrow image$ and $image \rightarrow speech$ queries.