A synthetic data set to benchmark anti-money laundering methods.

Bank transactions are highly confidential. As a result, there are no real public data sets that can be used to investigate and compare anti-money laundering (AML) methods in banks. This severely limits research on important AML problems such as efficiency, effectiveness, class imbalance, concept drift, and interpretability. To address the issue, we present SynthAML: a synthetic data set to benchmark statistical and machine learning methods for AML. The data set builds on real data from Spar Nord, a systemically important Danish bank, and contains 20,000 AML alerts and over 16 million transactions. Experimental results indicate that performance on SynthAML can be transferred to the real world. As use cases, we present and discuss open problems in the AML literature.

Automatic social distance estimation for photographic studies: Performance evaluation, test benchmark, and algorithm.

The social distancing regulations introduced to slow down the spread of COVID-19 virus directly affect a basic form of non-verbal communication, and there may be longer term impacts on human behavior and culture that remain to be analyzed in proxemics studies. To obtain quantitative results for such studies, large media and/or personal photo collections must be analyzed. Several social distance monitoring methods have been proposed for safety purposes, but they are not directly applicable to general photo collections with large variations in the imaging setup. In such studies, the interest shifts from safety to analyzing subtle differences in social distances. Currently, there is no suitable benchmark for developing such algorithms. Collecting images with measured ground-truth pair-wise distances using different camera settings is cumbersome. Moreover, performance evaluation for these algorithms is not straightforward, and there is no widely accepted evaluation protocol. In this paper, we provide an image dataset with measured pair-wise social distances under different camera positions and settings. We suggest a performance evaluation protocol and provide a benchmark to easily evaluate such algorithms. We also propose an automatic social distance estimation method that can be applied on general photo collections. Our method is a hybrid method that combines deep learning-based object detection and human pose estimation with projective geometry. The method can be applied on uncalibrated single images with known focal length and sensor size. The results on our benchmark are encouraging with 91% human detection rate and only 38.24% average relative distance estimation error among the detected people.

Single-layer vision transformers for more accurate early exits with less overhead.

Deploying deep learning models in time-critical applications with limited computational resources, for instance in edge computing systems and IoT networks, is a challenging task that often relies on dynamic inference methods such as early exiting. In this paper, we introduce a novel architecture for early exiting based on the vision transformer architecture, as well as a fine-tuning strategy that significantly increase the accuracy of early exit branches compared to conventional approaches while introducing less overhead. Through extensive experiments on image and audio classification as well as audiovisual crowd counting, we show that our method works for both classification and regression problems, and in both single- and multi-modal settings. Additionally, we introduce a novel method for integrating audio and visual modalities within early exits in audiovisual data analysis, that can lead to a more fine-grained dynamic inference.

Forecasting large-scale circulation regimes using deformable convolutional neural networks and global spatiotemporal climate data.

Classifying the state of the atmosphere into a finite number of large-scale circulation regimes is a popular way of investigating teleconnections, the predictability of severe weather events, and climate change. Here, we investigate a supervised machine learning approach based on deformable convolutional neural networks (deCNNs) and transfer learning to forecast the North Atlantic-European weather regimes during extended boreal winter for 1-15 days into the future. We apply state-of-the-art interpretation techniques from the machine learning literature to attribute particular regions of interest or potential teleconnections relevant for any given weather cluster prediction or regime transition. We demonstrate superior forecasting performance relative to several classical meteorological benchmarks, as well as logistic regression and random forests. Due to its wider field of view, we also observe deCNN achieving considerably better performance than regular convolutional neural networks at lead times beyond 5-6 days. Finally, we find transfer learning to be of paramount importance, similar to previous data-driven atmospheric forecasting studies.

Saliency-Based Multilabel Linear Discriminant Analysis.

Linear discriminant analysis (LDA) is a classical statistical machine-learning method, which aims to find a linear data transformation increasing class discrimination in an optimal discriminant subspace. Traditional LDA sets assumptions related to the Gaussian class distributions and single-label data annotations. In this article, we propose a new variant of LDA to be used in multilabel classification tasks for dimensionality reduction on original data to enhance the subsequent performance of any multilabel classifier. A probabilistic class saliency estimation approach is introduced for computing saliency-based weights for all instances. We use the weights to redefine the between-class and within-class scatter matrices needed for calculating the projection matrix. We formulate six different variants of the proposed saliency-based multilabel LDA (SMLDA) based on different prior information on the importance of each instance for their class(es) extracted from labels and features. Our experiments show that the proposed SMLDA leads to performance improvements in various multilabel classification problems compared to several competing dimensionality reduction methods.

Feedforward neural networks initialization based on discriminant learning.

In this paper, a novel data-driven method for weight initialization of Multilayer Perceptrons and Convolutional Neural Networks based on discriminant learning is proposed. The approach relaxes some of the limitations of competing data-driven methods, including unimodality assumptions, limitations on the architectures related to limited maximal dimensionalities of the corresponding projection spaces, as well as limitations related to high computational requirements due to the need of eigendecomposition on high-dimensional data. We also consider assumptions of the method on the data and propose a way to account for them in a form of a new normalization layer. The experiments on three large-scale image datasets show improved accuracy of the trained models compared to competing random-based and data-driven weight initialization methods, as well as better convergence properties in certain cases.

Supervised Domain Adaptation: A Graph Embedding Perspective and a Rectified Experimental Protocol.

Domain Adaptation is the process of alleviating distribution gaps between data from different domains. In this paper, we show that Domain Adaptation methods using pair-wise relationships between source and target domain data can be formulated as a Graph Embedding in which the domain labels are incorporated into the structure of the intrinsic and penalty graphs. Specifically, we analyse the loss functions of three existing state-of-the-art Supervised Domain Adaptation methods and demonstrate that they perform Graph Embedding. Moreover, we highlight some generalisation and reproducibility issues related to the experimental setup commonly used to demonstrate the few-shot learning capabilities of these methods. To assess and compare Supervised Domain Adaptation methods accurately, we propose a rectified evaluation protocol, and report updated benchmarks on the standard datasets Office31 (Amazon, DSLR, and Webcam), Digits (MNIST, USPS, SVHN, and MNIST-M) and VisDA (Synthetic, Real).

Self-organized Operational Neural Networks with Generative Neurons.

Operational Neural Networks (ONNs) have recently been proposed to address the well-known limitations and drawbacks of conventional Convolutional Neural Networks (CNNs) such as network homogeneity with the sole linear neuron model. ONNs are heterogeneous networks with a generalized neuron model. However the operator search method in ONNs is not only computationally demanding, but the network heterogeneity is also limited since the same set of operators will then be used for all neurons in each layer. Moreover, the performance of ONNs directly depends on the operator set library used, which introduces a certain risk of performance degradation especially when the optimal operator set required for a particular task is missing from the library. In order to address these issues and achieve an ultimate heterogeneity level to boost the network diversity along with computational efficiency, in this study we propose Self-organized ONNs (Self-ONNs) with generative neurons that can adapt (optimize) the nodal operator of each connection during the training process. Moreover, this ability voids the need of having a fixed operator set library and the prior operator search within the library in order to find the best possible set of operators. We further formulate the training method to back-propagate the error through the operational layers of Self-ONNs. Experimental results over four challenging problems demonstrate the superior learning capability and computational efficiency of Self-ONNs over conventional ONNs and CNNs.

Deep learning and computer vision will transform entomology.

Most animal species on Earth are insects, and recent reports suggest that their abundance is in drastic decline. Although these reports come from a wide range of insect taxa and regions, the evidence to assess the extent of the phenomenon is sparse. Insect populations are challenging to study, and most monitoring methods are labor intensive and inefficient. Advances in computer vision and deep learning provide potential new solutions to this global challenge. Cameras and other sensors can effectively, continuously, and noninvasively perform entomological observations throughout diurnal and seasonal cycles. The physical appearance of specimens can also be captured by automated imaging in the laboratory. When trained on these data, deep learning models can provide estimates of insect abundance, biomass, and diversity. Further, deep learning models can quantify variation in phenotypic traits, behavior, and interactions. Here, we connect recent developments in deep learning and computer vision to the urgent demand for more cost-efficient monitoring of insects and other invertebrates. We present examples of sensor-based monitoring of insects. We show how deep learning tools can be applied to exceptionally large datasets to derive ecological information and discuss the challenges that lie ahead for the implementation of such solutions in entomology. We identify four focal areas, which will facilitate this transformation: 1) validation of image-based taxonomic identification; 2) generation of sufficient training data; 3) development of public, curated reference databases; and 4) solutions to integrate deep learning and molecular tools.

Multilinear Compressive Learning.

Compressive learning (CL) is an emerging topic that combines signal acquisition via compressive sensing (CS) and machine learning to perform inference tasks directly on a small number of measurements. Many data modalities naturally have a multidimensional or tensorial format, with each dimension or tensor mode representing different features such as the spatial and temporal information in video sequences or the spatial and spectral information in hyperspectral images. However, in existing CL frameworks, the CS component utilizes either random or learned linear projection on the vectorized signal to perform signal acquisition, thus discarding the multidimensional structure of the signals. In this article, we propose multilinear CL (MCL), a framework that takes into account the tensorial nature of multidimensional signals in the acquisition step and builds the subsequent inference model on the structurally sensed measurements. Our theoretical complexity analysis shows that the proposed framework is more efficient compared to its vector-based counterpart in both memory and computation requirement. With extensive experiments, we also empirically show that our MCL framework outperforms the vector-based framework in object classification and face recognition tasks, and scales favorably when the dimensionalities of the original signals increase, making it highly efficient for high-dimensional multidimensional signals.

Hypersphere-Based Weight Imprinting for Few-Shot Learning on Embedded Devices.

Weight imprinting (WI) was recently introduced as a way to perform gradient descent-free few-shot learning. Due to this, WI was almost immediately adapted for performing few-shot learning on embedded neural network accelerators that do not support back-propagation, e.g., edge tensor processing units. However, WI suffers from many limitations, e.g., it cannot handle novel categories with multimodal distributions and special care should be given to avoid overfitting the learned embeddings on the training classes since this can have a devastating effect on classification accuracy (for the novel categories). In this article, we propose a novel hypersphere-based WI approach that is capable of training neural networks in a regularized, imprinting-aware way effectively overcoming the aforementioned limitations. The effectiveness of the proposed method is demonstrated using extensive experiments on three image data sets.

Mid-price prediction based on machine learning methods with technical and quantitative indicators.

Stock price prediction is a challenging task, in which machine learning methods have recently been successfully used. In this paper, we extract over 270 hand-crafted features (factors) inspired by technical indicators and quantitative analysis and test their validity on short-term mid-price movement prediction for Nordic TotalView-ITCH stocks. The suggested feature list represents one of the most extensive studies in the field of financial feature engineering. We focus on a wrapper feature selection method using entropy, least-mean squares, and linear discriminant analysis. We also introduce a novel quantitative feature based on adaptive logistic regression for online learning. The proposed feature is consistently selected as the first feature among a large number of indicators used in this study. We further examine the best combinations of features using a high-frequency limit order book Nordic database. Our results suggest that sorting methods and classifiers can be used in such a way that one can reach the best classification performance with a combination of only a few advanced hand-crafted features.

Species-level image classification with convolutional neural network enables insect identification from habitus images.

Changes in insect biomass, abundance, and diversity are challenging to track at sufficient spatial, temporal, and taxonomic resolution. Camera traps can capture habitus images of ground-dwelling insects. However, currently sampling involves manually detecting and identifying specimens. Here, we test whether a convolutional neural network (CNN) can classify habitus images of ground beetles to species level, and estimate how correct classification relates to body size, number of species inside genera, and species identity.We created an image database of 65,841 museum specimens comprising 361 carabid beetle species from the British Isles and fine-tuned the parameters of a pretrained CNN from a training dataset. By summing up class confidence values within genus, tribe, and subfamily and setting a confidence threshold, we trade-off between classification accuracy, precision, and recall and taxonomic resolution.The CNN classified 51.9% of 19,164 test images correctly to species level and 74.9% to genus level. Average classification recall on species level was 50.7%. Applying a threshold of 0.5 increased the average classification recall to 74.6% at the expense of taxonomic resolution. Higher top value from the output layer and larger sized species were more often classified correctly, as were images of species in genera with few species.Fine-tuning enabled us to classify images with a high mean recall for the whole test dataset to species or higher taxonomic levels, however, with high variability. This indicates that some species are more difficult to identify because of properties such as their body size or the number of related species.Together, species-level image classification of arthropods from museum collections and ecological monitoring can substantially increase the amount of occurrence data that can feasibly be collected. These tools thus provide new opportunities in understanding and predicting ecological responses to environmental change.

Heterogeneous Multilayer Generalized Operational Perceptron.

The traditional multilayer perceptron (MLP) using a McCulloch-Pitts neuron model is inherently limited to a set of neuronal activities, i.e., linear weighted sum followed by nonlinear thresholding step. Previously, generalized operational perceptron (GOP) was proposed to extend the conventional perceptron model by defining a diverse set of neuronal activities to imitate a generalized model of biological neurons. Together with GOP, a progressive operational perceptron (POP) algorithm was proposed to optimize a predefined template of multiple homogeneous layers in a layerwise manner. In this paper, we propose an efficient algorithm to learn a compact, fully heterogeneous multilayer network that allows each individual neuron, regardless of the layer, to have distinct characteristics. Based on the complexity of the problem, the proposed algorithm operates in a progressive manner on a neuronal level, searching for a compact topology, not only in terms of depth but also width, i.e., the number of neurons in each layer. The proposed algorithm is shown to outperform other related learning methods in extensive experiments on several classification problems.

Deep Adaptive Input Normalization for Time Series Forecasting.

Deep learning (DL) models can be used to tackle time series analysis tasks with great success. However, the performance of DL models can degenerate rapidly if the data are not appropriately normalized. This issue is even more apparent when DL is used for financial time series forecasting tasks, where the nonstationary and multimodal nature of the data pose significant challenges and severely affect the performance of DL models. In this brief, a simple, yet effective, neural layer that is capable of adaptively normalizing the input time series, while taking into account the distribution of the data, is proposed. The proposed layer is trained in an end-to-end fashion using backpropagation and leads to significant performance improvements compared to other evaluated normalization schemes. The proposed method differs from traditional normalization methods since it learns how to perform normalization for a given task instead of using a fixed normalization scheme. At the same time, it can be directly applied to any new time series without requiring retraining. The effectiveness of the proposed method is demonstrated using a large-scale limit order book data set, as well as a load forecasting data set.

Temporal Attention-Augmented Bilinear Network for Financial Time-Series Data Analysis.

Financial time-series forecasting has long been a challenging problem because of the inherently noisy and stochastic nature of the market. In the high-frequency trading, forecasting for trading purposes is even a more challenging task, since an automated inference system is required to be both accurate and fast. In this paper, we propose a neural network layer architecture that incorporates the idea of bilinear projection as well as an attention mechanism that enables the layer to detect and focus on crucial temporal information. The resulting network is highly interpretable, given its ability to highlight the importance and contribution of each temporal instance, thus allowing further analysis on the time instances of interest. Our experiments in a large-scale limit order book data set show that a two-hidden-layer network utilizing our proposed layer outperforms by a large margin all existing state-of-the-art results coming from much deeper architectures while requiring far fewer computations.

Improving efficiency in convolutional neural networks with multilinear filters.

The excellent performance of deep neural networks has enabled us to solve several automatization problems, opening an era of autonomous devices. However, current deep net architectures are heavy with millions of parameters and require billions of floating point operations. Several works have been developed to compress a pre-trained deep network to reduce memory footprint and, possibly, computation. Instead of compressing a pre-trained network, in this work, we propose a generic neural network layer structure employing multilinear projection as the primary feature extractor. The proposed architecture requires several times less memory as compared to the traditional Convolutional Neural Networks (CNN), while inherits the similar design principles of a CNN. In addition, the proposed architecture is equipped with two computation schemes that enable computation reduction or scalability. Experimental results show the effectiveness of our compact projection that outperforms traditional CNN, while requiring far fewer parameters.

Generalized Multi-View Embedding for Visual Recognition and Cross-Modal Retrieval.

In this paper, the problem of multi-view embedding from different visual cues and modalities is considered. We propose a unified solution for subspace learning methods using the Rayleigh quotient, which is extensible for multiple views, supervised learning, and nonlinear embeddings. Numerous methods including canonical correlation analysis, partial least square regression, and linear discriminant analysis are studied using specific intrinsic and penalty graphs within the same framework. Nonlinear extensions based on kernels and (deep) neural networks are derived, achieving better performance than the linear ones. Moreover, a novel multi-view modular discriminant analysis is proposed by taking the view difference into consideration. We demonstrate the effectiveness of the proposed multi-view embedding methods on visual object recognition and cross-modal image retrieval, and obtain superior results in both applications compared to related methods.

Class-Specific Kernel Discriminant Analysis Revisited: Further Analysis and Extensions.

In this paper, we revisit class-specific kernel discriminant analysis (KDA) formulation, which has been applied in various problems, such as human face verification and human action recognition. We show that the original optimization problem solved for the determination of class-specific discriminant projections is equivalent to a low-rank kernel regression (LRKR) problem using training data-independent target vectors. In addition, we show that the regularized version of class-specific KDA is equivalent to a regularized LRKR problem, exploiting the same targets. This analysis allows us to devise a novel fast solution. Furthermore, we devise novel incremental, approximate and deep (hierarchical) variants. The proposed methods are tested in human facial image and action video verification problems, where their effectiveness and efficiency is shown.

Graph Embedded Extreme Learning Machine.

In this paper, we propose a novel extension of the extreme learning machine (ELM) algorithm for single-hidden layer feedforward neural network training that is able to incorporate subspace learning (SL) criteria on the optimization process followed for the calculation of the network's output weights. The proposed graph embedded ELM (GEELM) algorithm is able to naturally exploit both intrinsic and penalty SL criteria that have been (or will be) designed under the graph embedding framework. In addition, we extend the proposed GEELM algorithm in order to be able to exploit SL criteria in arbitrary (even infinite) dimensional ELM spaces. We evaluate the proposed approach on eight standard classification problems and nine publicly available datasets designed for three problems related to human behavior analysis, i.e., the recognition of human face, facial expression, and activity. Experimental results denote the effectiveness of the proposed approach, since it outperforms other ELM-based classification schemes in all the cases.