Since training deep neural networks takes significant computational resources, extending the training dataset with new data is difficult, as it typically requires complete retraining. Moreover, specific applications do not allow costly retraining due to time or computational constraints. We address this issue by proposing a novel Bayesian update method for deep neural networks by using a last-layer Laplace approximation. Concretely, we leverage second-order optimization techniques on the Gaussian posterior distribution of a Laplace approximation, computing the inverse Hessian matrix in closed form. This way, our method allows for fast and effective updates upon the arrival of new data in a stationary setting. A large-scale evaluation study across different data modalities confirms that our updates are a fast and competitive alternative to costly retraining. Furthermore, we demonstrate its applicability in a deep active learning scenario by using our update to improve existing selection strategies.
In modern industrial systems, diagnosing faults in time and using the best methods becomes more and more crucial. It is possible to fail a system or to waste resources if faults are not detected or are detected late. Machine learning and deep learning have proposed various methods for data-based fault diagnosis, and we are looking for the most reliable and practical ones. This paper aims to develop a framework based on deep learning and reinforcement learning for fault detection. We can increase accuracy, overcome data imbalance, and better predict future defects by updating the reinforcement learning policy when new data is received. By implementing this method, we will see an increase of $3\%$ in all evaluation metrics, an improvement in prediction speed, and $3\%$ - $4\%$ in all evaluation metrics compared to typical backpropagation multi-layer neural network prediction with similar parameters.
Stress is prevalent in many aspects of everyday life including work, healthcare, and social interactions. Many works have studied handcrafted features from various bio-signals that are indicators of stress. Recently, deep learning models have also been proposed to detect stress. Typically, stress models are trained and validated on the same dataset, often involving one stressful scenario. However, it is not practical to collect stress data for every scenario. So, it is crucial to study the generalizability of these models and determine to what extent they can be used in other scenarios. In this paper, we explore the generalization capabilities of Electrocardiogram (ECG)-based deep learning models and models based on handcrafted ECG features, i.e., Heart Rate Variability (HRV) features. To this end, we train three HRV models and two deep learning models that use ECG signals as input. We use ECG signals from two popular stress datasets - WESAD and SWELL-KW - differing in terms of stressors and recording devices. First, we evaluate the models using leave-one-subject-out (LOSO) cross-validation using training and validation samples from the same dataset. Next, we perform a cross-dataset validation of the models, that is, LOSO models trained on the WESAD dataset are validated using SWELL-KW samples and vice versa. While deep learning models achieve the best results on the same dataset, models based on HRV features considerably outperform them on data from a different dataset. This trend is observed for all the models on both datasets. Therefore, HRV models are a better choice for stress recognition in applications that are different from the dataset scenario. To the best of our knowledge, this is the first work to compare the cross-dataset generalizability between ECG-based deep learning models and HRV models.
In this survey, we provide a comprehensive review of more than 200 papers, technical reports, and GitHub repositories published over the last 10 years on the recent developments of deep learning techniques for iris recognition, covering broad topics on algorithm designs, open-source tools, open challenges, and emerging research. First, we conduct a comprehensive analysis of deep learning techniques developed for two main sub-tasks in iris biometrics: segmentation and recognition. Second, we focus on deep learning techniques for the robustness of iris recognition systems against presentation attacks and via human-machine pairing. Third, we delve deep into deep learning techniques for forensic application, especially in post-mortem iris recognition. Fourth, we review open-source resources and tools in deep learning techniques for iris recognition. Finally, we highlight the technical challenges, emerging research trends, and outlook for the future of deep learning in iris recognition.
Spatially resolved transcriptomics (SRT) has evolved rapidly through various technologies, enabling scientists to investigate both morphological contexts and gene expression profiling at single-cell resolution in parallel. SRT data are complex and multi-modal, comprising gene expression matrices, spatial information, and often high-resolution histology images. Because of this complexity and multi-modality, sophisticated computational algorithms are required to accurately analyze SRT data. Most efforts in this domain have been made to utilize conventional machine learning and statistical approaches, exhibiting sub-optimal results due to the complicated nature of SRT datasets. To address these shortcomings, researchers have recently employed deep learning algorithms including various state-of-the-art methods mainly in spatial clustering, spatially variable gene identification, and alignment. While great progress has been made in developing deep learning-based models for SRT data analysis, further improvement is still needed to create more biologically aware models that consider aspects such as phylogeny-aware clustering or the analysis of small histology image patches. Additionally, strategies for batch effect removal, normalization, and handling overdispersion and zero inflation patterns of gene expression are still needed in the analysis of SRT data using deep learning methods. In this paper, we provide a comprehensive overview of these deep learning methods, including their strengths and limitations. We also highlight new frontiers, current challenges, limitations, and open questions in this field. Also, we provide a comprehensive list of all available SRT databases that can be used as an extensive resource for future studies.
Optical spectroscopy is indispensable for research and development in nanoscience and nanotechnology, microelectronics, energy, and advanced manufacturing. Advanced optical spectroscopy tools often require both specifically designed high-end instrumentation and intricate data analysis techniques. Beyond the common analytical tools, deep learning methods are well suited for interpreting high-dimensional and complicated spectroscopy data. They offer great opportunities to extract subtle and deep information about optical properties of materials with simpler optical setups, which would otherwise require sophisticated instrumentation. In this work, we propose a computational ellipsometry approach based on a conventional tabletop optical microscope and a deep learning model called EllipsoNet. Without any prior knowledge about the multilayer substrates, EllipsoNet can predict the complex refractive indices of thin films on top of these nontrivial substrates from experimentally measured optical reflectance spectra with high accuracies. This task was not feasible previously with traditional reflectometry or ellipsometry methods. Fundamental physical principles, such as the Kramers-Kronig relations, are spontaneously learned by the model without any further training. This approach enables in-operando optical characterization of functional materials within complex photonic structures or optoelectronic devices.
The superior performance of deep learning relies heavily on a large collection of sample data, but the data insufficiency problem turns out to be relatively common in global electricity markets. How to prevent overfitting in this case becomes a fundamental challenge when training deep learning models in different market applications. With this in mind, we propose a general framework, namely Knowledge-Augmented Training (KAT), to improve the sample efficiency, and the main idea is to incorporate domain knowledge into the training procedures of deep learning models. Specifically, we propose a novel data augmentation technique to generate some synthetic data, which are later processed by an improved training strategy. This KAT methodology follows and realizes the idea of combining analytical and deep learning models together. Modern learning theories demonstrate the effectiveness of our method in terms of effective prediction error feedbacks, a reliable loss function, and rich gradient noises. At last, we study two popular applications in detail: user modeling and probabilistic price forecasting. The proposed method outperforms other competitors in all numerical tests, and the underlying reasons are explained by further statistical and visualization results.
Earth observation (EO), aiming at monitoring the state of planet Earth using remote sensing data, is critical for improving our daily lives and living environment. With a growing number of satellites in orbit, an increasing number of datasets with diverse sensors and research domains are being published to facilitate the research of the remote sensing community. This paper presents a comprehensive review of more than 500 publicly published datasets, including research domains like agriculture, land use and land cover, disaster monitoring, scene understanding, vision-language models, foundation models, climate change, and weather forecasting. We systematically analyze these EO datasets from four aspects: volume, resolution distributions, research domains, and the correlation between datasets. Based on the dataset attributes, we propose to measure, rank, and select datasets to build a new benchmark for model evaluation. Furthermore, a new platform for EO, termed EarthNets, is released to achieve a fair and consistent evaluation of deep learning methods on remote sensing data. EarthNets supports standard dataset libraries and cutting-edge deep learning models to bridge the gap between the remote sensing and machine learning communities. Based on this platform, extensive deep-learning methods are evaluated on the new benchmark. The insightful results are beneficial to future research. The platform and dataset collections are publicly available at https://earthnets.github.io.
The Retinex model is one of the most representative and effective methods for low-light image enhancement. However, the Retinex model does not explicitly tackle the noise problem, and shows unsatisfactory enhancing results. In recent years, due to the excellent performance, deep learning models have been widely used in low-light image enhancement. However, these methods have two limitations: i) The desirable performance can only be achieved by deep learning when a large number of labeled data are available. However, it is not easy to curate massive low/normal-light paired data; ii) Deep learning is notoriously a black-box model [1]. It is difficult to explain their inner-working mechanism and understand their behaviors. In this paper, using a sequential Retinex decomposition strategy, we design a plug-and-play framework based on the Retinex theory for simultaneously image enhancement and noise removal. Meanwhile, we develop a convolutional neural network-based (CNN-based) denoiser into our proposed plug-and-play framework to generate a reflectance component. The final enhanced image is produced by integrating the illumination and reflectance with gamma correction. The proposed plug-and-play framework can facilitate both post hoc and ad hoc interpretability. Extensive experiments on different datasets demonstrate that our framework outcompetes the state-of-the-art methods in both image enhancement and denoising.
In spite of years of research, the mechanisms that underlie the development of schizophrenia, as well as its relapse, symptomatology, and treatment, continue to be a mystery. The absence of appropriate analytic tools to deal with the variable and complicated nature of schizophrenia may be one of the factors that contribute to the development of this disorder. Deep learning is a subfield of artificial intelligence that was inspired by the nervous system. In recent years, deep learning has made it easier to model and analyse complicated, high-dimensional, and nonlinear systems. Research on schizophrenia is one of the many areas of study that has been revolutionised as a result of the outstanding accuracy that deep learning algorithms have demonstrated in classification and prediction tasks. Deep learning has the potential to become a powerful tool for understanding the mechanisms that are at the root of schizophrenia. In addition, a growing variety of techniques aimed at improving model interpretability and causal reasoning are contributing to this trend. Using multi-site fMRI data and a variety of deep learning approaches, this study seeks to identify different types of schizophrenia. Our proposed method of temporal aggregation of the 4D fMRI data outperforms existing work. In addition, this study aims to shed light on the strength of connections between various brain areas in schizophrenia individuals.
Deep learning model compression is an improving and important field for the edge deployment of deep learning models. Given the increasing size of the models and their corresponding power consumption, it is vital to decrease the model size and compute requirement without a significant drop in the model's performance. In this paper, we present model compression algorithms for both non-retraining and retraining conditions. In the first case where retraining of the model is not feasible due to lack of access to the original data or absence of necessary compute resources while only having access to off-the-shelf models, we propose the Bin & Quant algorithm for compression of the deep learning models using the sensitivity of the network parameters. This results in 13x compression of the speech command and control model and 7x compression of the DeepSpeech2 models. In the second case when the models can be retrained and utmost compression is required for the negligible loss in accuracy, we propose our novel gradient-weighted k-means clustering algorithm (GWK). This method uses the gradients in identifying the important weight values in a given cluster and nudges the centroid towards those values, thereby giving importance to sensitive weights. Our method effectively combines product quantization with the EWGS[1] algorithm for sub-1-bit representation of the quantized models. We test our GWK algorithm on the CIFAR10 dataset across a range of models such as ResNet20, ResNet56, MobileNetv2 and show 35x compression on quantized models for less than 2% absolute loss in accuracy compared to the floating-point models.
While domain-specific data augmentation can be useful in training neural networks for medical imaging tasks, such techniques have not been widely used to date. Here, we test whether domain-specific data augmentation is useful for medical imaging using a well-benchmarked task: view classification on fetal ultrasound FETAL-125 and OB-125 datasets. We found that using a context-preserving cut-paste strategy, we could create valid training data as measured by performance of the resulting trained model on the benchmark test dataset. When used in an online fashion, models trained on this data performed similarly to those trained using traditional data augmentation (FETAL-125 F-score 85.33+/-0.24 vs 86.89+/-0.60, p-value 0.0139; OB-125 F-score 74.60+/-0.11 vs 72.43+/-0.62, p-value 0.0039). Furthermore, the ability to perform augmentations during training time, as well as the ability to apply chosen augmentations equally across data classes, are important considerations in designing a bespoke data augmentation. Finally, we provide open-source code to facilitate running bespoke data augmentations in an online fashion. Taken together, this work expands the ability to design and apply domain-guided data augmentations for medical imaging tasks.
Raster well-log images are digital representations of well-logs data generated over the years. Raster digital well logs represent bitmaps of the log image in a rectangular array of black (zeros) and white dots (ones) called pixels. Experts study the raster logs manually or with software applications that still require a tremendous amount of manual input. Besides the loss of thousands of person-hours, this process is erroneous and tedious. To digitize these raster logs, one must buy a costly digitizer that is not only manual and time-consuming but also a hidden technical debt since enterprises stand to lose more money in additional servicing and consulting charges. We propose a deep neural network architecture called VeerNet to semantically segment the raster images from the background grid and classify and digitize the well-log curves. Raster logs have a substantially greater resolution than images traditionally consumed by image segmentation pipelines. Since the input has a low signal-to-resolution ratio, we require rapid downsampling to alleviate unnecessary computation. We thus employ a modified UNet-inspired architecture that balances retaining key signals and reducing result dimensionality. We use attention augmented read-process-write architecture. This architecture efficiently classifies and digitizes the curves with an overall F1 score of 35% and IoU of 30%. When compared to the actual las values for Gamma-ray and derived value of Gamma-ray from VeerNet, a high Pearson coefficient score of 0.62 was achieved.
Predicting the occurence of rare and extreme events in complex systems is a well-known problem in non-equilibrium physics. These events can have huge impacts on human societies. New approaches have emerged in the last ten years, which better estimate tail distributions. They often use large deviation concepts without the need to perform heavy direct ensemble simulations. In particular, a well-known approach is to derive a minimum action principle and to find its minimizers. The analysis of rare reactive events in non-equilibrium systems without detailed balance is notoriously difficult either theoretically and computationally. They are described in the limit of small noise by the Freidlin-Wentzell action. We propose here a new method which minimizes the geometrical action instead using neural networks: it is called deep gMAM. It relies on a natural and simple machine-learning formulation of the classical gMAM approach. We give a detailed description of the method as well as many examples. These include bimodal switches in complex stochastic (partial) differential equations, quasi-potential estimates, and extreme events in Burgers turbulence.
Training large neural networks with meaningful/usable differential privacy security guarantees is a demanding challenge. In this paper, we tackle this problem by revisiting the two key operations in Differentially Private Stochastic Gradient Descent (DP-SGD): 1) iterative perturbation and 2) gradient clipping. We propose a generic optimization framework, called {\em ModelMix}, which performs random aggregation of intermediate model states. It strengthens the composite privacy analysis utilizing the entropy of the training trajectory and improves the $(\epsilon, \delta)$ DP security parameters by an order of magnitude. We provide rigorous analyses for both the utility guarantees and privacy amplification of ModelMix. In particular, we present a formal study on the effect of gradient clipping in DP-SGD, which provides theoretical instruction on how hyper-parameters should be selected. We also introduce a refined gradient clipping method, which can further sharpen the privacy loss in private learning when combined with ModelMix. Thorough experiments with significant privacy/utility improvement are presented to support our theory. We train a Resnet-20 network on CIFAR10 with $70.4\%$ accuracy via ModelMix given $(\epsilon=8, \delta=10^{-5})$ DP-budget, compared to the same performance but with $(\epsilon=145.8,\delta=10^{-5})$ using regular DP-SGD; assisted with additional public low-dimensional gradient embedding, one can further improve the accuracy to $79.1\%$ with $(\epsilon=6.1, \delta=10^{-5})$ DP-budget, compared to the same performance but with $(\epsilon=111.2, \delta=10^{-5})$ without ModelMix.
Deep learning (DL) has been widely adopted those last years but they are computing-intensive method. Therefore, scientists proposed diverse optimization to accelerate their predictions for end-user applications. However, no single inference framework currently dominates in terms of performance. This paper takes a holistic approach to conduct an empirical comparison and analysis of four representative DL inference frameworks. First, given a selection of CPU-GPU configurations, we show that for a specific DL framework, different configurations of its settings may have a significant impact on the prediction speed, memory, and computing power. Second, to the best of our knowledge, this study is the first to identify the opportunities for accelerating the ensemble of co-localized models in the same GPU. This measurement study provides an in-depth empirical comparison and analysis of four representative DL frameworks and offers practical guidance for service providers to deploy and deliver DL predictions.
Causal mediation analysis can unpack the black box of causality and is therefore a powerful tool for disentangling causal pathways in biomedical and social sciences, and also for evaluating machine learning fairness. To reduce bias for estimating Natural Direct and Indirect Effects in mediation analysis, we propose a new method called DeepMed that uses deep neural networks (DNNs) to cross-fit the infinite-dimensional nuisance functions in the efficient influence functions. We obtain novel theoretical results that our DeepMed method (1) can achieve semiparametric efficiency bound without imposing sparsity constraints on the DNN architecture and (2) can adapt to certain low dimensional structures of the nuisance functions, significantly advancing the existing literature on DNN-based semiparametric causal inference. Extensive synthetic experiments are conducted to support our findings and also expose the gap between theory and practice. As a proof of concept, we apply DeepMed to analyze two real datasets on machine learning fairness and reach conclusions consistent with previous findings.
It is often said that a deep learning model is "invariant" to some specific type of transformation. However, what is meant by this statement strongly depends on the context in which it is made. In this paper we explore the nature of invariance and equivariance of deep learning models with the goal of better understanding the ways in which they actually capture these concepts on a formal level. We introduce a family of invariance and equivariance metrics that allows us to quantify these properties in a way that disentangles them from other metrics such as loss or accuracy. We use our metrics to better understand the two most popular methods used to build invariance into networks: data augmentation and equivariant layers. We draw a range of conclusions about invariance and equivariance in deep learning models, ranging from whether initializing a model with pretrained weights has an effect on a trained model's invariance, to the extent to which invariance learned via training can generalize to out-of-distribution data.
The deep learning (DL)-based methods of low-level tasks have many advantages over the traditional camera in terms of hardware prospects, error accumulation and imaging effects. Recently, the application of deep learning to replace the image signal processing (ISP) pipeline has appeared one after another; however, there is still a long way to go towards real landing. In this paper, we show the possibility of learning-based method to achieve real-time high-performance processing in the ISP pipeline. We propose LW-ISP, a novel architecture designed to implicitly learn the image mapping from RAW data to RGB image. Based on U-Net architecture, we propose the fine-grained attention module and a plug-and-play upsampling block suitable for low-level tasks. In particular, we design a heterogeneous distillation algorithm to distill the implicit features and reconstruction information of the clean image, so as to guide the learning of the student model. Our experiments demonstrate that LW-ISP has achieved a 0.38 dB improvement in PSNR compared to the previous best method, while the model parameters and calculation have been reduced by 23 times and 81 times. The inference efficiency has been accelerated by at least 15 times. Without bells and whistles, LW-ISP has achieved quite competitive results in ISP subtasks including image denoising and enhancement.
When logos are increasingly created, logo detection has gradually become a research hotspot across many domains and tasks. Recent advances in this area are dominated by deep learning-based solutions, where many datasets, learning strategies, network architectures, etc. have been employed. This paper reviews the advance in applying deep learning techniques to logo detection. Firstly, we discuss a comprehensive account of public datasets designed to facilitate performance evaluation of logo detection algorithms, which tend to be more diverse, more challenging, and more reflective of real life. Next, we perform an in-depth analysis of the existing logo detection strategies and the strengths and weaknesses of each learning strategy. Subsequently, we summarize the applications of logo detection in various fields, from intelligent transportation and brand monitoring to copyright and trademark compliance. Finally, we analyze the potential challenges and present the future directions for the development of logo detection to complete this survey.