Deep Learning-Based Computer Vision Application with Multiple Built-In Data Science-Oriented Capabilities

Deep learning (DL) model training must address the memory bottleneck to continue scaling. Processing-in-memory approaches can be a viable solution as they move computations near or into the memory, reducing substantial data movement. However, to deploy applications on such hardware, end-to-end software support is crucial for efficient computation mapping and scheduling as well as extensible code generation, but no consideration has been made for DL training workloads. In this paper, we propose XLA-NDP, a compiler and runtime solution for NDPX, a near-data processing (NDP) architecture integrated with an existing DL training framework. XLA-NDP offloads NDPX kernels and schedules them to overlap with GPU kernels to maximize parallelism based on GPU and NDPX costs, while providing a template-based code generator with low-level optimizations. The experiments showed that XLA-NDP provides up to 1.41x speedup (1.24x on average) over the GPU baseline for four DL model training.

Artificial intelligence (AI) has the ability to predict rheological properties and constituent composition of 3D-printed materials with appropriately trained models. However, these models are not currently available for use. In this work, we trained deep learning (DL) models to (1) predict the rheological properties, such as the storage (G’) and loss (G”) moduli, of 3D-printed polyacrylamide (PAA) substrates, and (2) predict the composition of materials and associated 3D printing parameters for a desired pair of G’ and G”. We employed a multilayer perceptron (MLP) and successfully predicted G’ and G” from seven gel constituent parameters in a multivariate regression process. We used a grid-search algorithm along with 10-fold cross validation to tune the hyperparameters of the MLP, and found the R2 value to be 0.89. Next, we adopted two generative DL models named variational autoencoder (VAE) and conditional variational autoencoder (CVAE) to learn data patterns and generate constituent compositions. With these generative models, we produced synthetic data with the same statistical distribution as the real data of actual hydrogel fabrication, which was then validated using Student’s t-test and an autoencoder (AE) anomaly detector. We found that none of the seven generated gel constituents were significantly different from the real data. Our trained DL models were successful in mapping the input–output relationship for the 3D-printed hydrogel substrates, which can predict multiple variables from a handful of input variables and vice versa.

PurposeTo conduct a head-to-head comparison between deep learning (DL) and radiomics models across institutions for predicting microvascular invasion (MVI) in hepatocellular carcinoma (HCC) and to investigate the model robustness and generalizability through rigorous internal and external validation.MethodsThis retrospective study included 2304 preoperative images of 576 HCC lesions from two centers, with MVI status determined by postoperative histopathology. We developed DL and radiomics models for predicting the presence of MVI using B-mode ultrasound, contrast-enhanced ultrasound (CEUS) at the arterial, portal, and delayed phases, and a combined modality (B + CEUS). For radiomics, we constructed models with enlarged vs. original regions of interest (ROIs). A cross-validation approach was performed by training models on one center’s dataset and validating the other, and vice versa. This allowed assessment of the validity of different ultrasound modalities and the cross-center robustness of the models. The optimal model combined with alpha-fetoprotein (AFP) was also validated. The head-to-head comparison was based on the area under the receiver operating characteristic curve (AUC).ResultsThirteen DL models and 25 radiomics models using different ultrasound modalities were constructed and compared. B + CEUS was the optimal modality for both DL and radiomics models. The DL model achieved AUCs of 0.802–0.818 internally and 0.667–0.688 externally across the two centers, whereas radiomics achieved AUCs of 0.749–0.869 internally and 0.646–0.697 externally. The radiomics models showed overall improvement with enlarged ROIs (P < 0.05 for both CEUS and B + CEUS modalities). The DL models showed good cross-institutional robustness (P > 0.05 for all modalities, 1.6–2.1% differences in AUC for the optimal modality), whereas the radiomics models had relatively limited robustness across the two centers (12% drop-off in AUC for the optimal modality). Adding AFP improved the DL models (P < 0.05 externally) and well maintained the robustness, but did not benefit the radiomics model (P > 0.05).ConclusionCross-institutional validation indicated that DL demonstrated better robustness than radiomics for preoperative MVI prediction in patients with HCC, representing a promising solution to non-standardized ultrasound examination procedures.

PurposeTo develop generative adversarial networks (GANs) that synthesize realistic anterior segment optical coherence tomography (AS-OCT) images and evaluate deep learning (DL) models that are trained on real and synthetic datasets for detecting angle closure.MethodsThe GAN architecture was adopted and trained on the dataset with AS-OCT images collected from the Joint Shantou International Eye Center of Shantou University and the Chinese University of Hong Kong, synthesizing open- and closed-angle AS-OCT images. A visual Turing test with two glaucoma specialists was performed to assess the image quality of real and synthetic images. DL models, trained on either real or synthetic datasets, were developed. Using the clinicians’ grading of the AS-OCT images as the reference standard, we compared the diagnostic performance of open-angle vs. closed-angle detection of DL models and the AS-OCT parameter, defined as a trabecular-iris space area 750 µm anterior to the scleral spur (TISA750), in a small independent validation dataset.ResultsThe GAN training included 28,643 AS-OCT anterior chamber angle (ACA) images. The real and synthetic datasets for DL model training have an equal distribution of open- and closed-angle images (all with 10,000 images each). The independent validation dataset included 238 open-angle and 243 closed-angle AS-OCT ACA images. The image quality of real versus synthetic AS-OCT images was similar, as assessed by the two glaucoma specialists, except for the scleral spur visibility. For the independent validation dataset, both DL models achieved higher areas under the curve compared with TISA750. Two DL models had areas under the curve of 0.97 (95% confidence interval, 0.96–0.99) and 0.94 (95% confidence interval, 0.92–0.96).ConclusionsThe GAN synthetic AS-OCT images appeared to be of good quality, according to the glaucoma specialists. The DL models, trained on all-synthetic AS-OCT images, can achieve high diagnostic performance.Translational RelevanceThe GANs can generate realistic AS-OCT images, which can also be used to train DL models.

Electricity load forecasting is a crucial undertaking within all the deregulated markets globally. Among the research challenges on a global scale, the investigation of deep transfer learning (DTL) in the field of electricity load forecasting represents a fundamental effort that can inform artificial intelligence applications in general. In this paper, a comprehensive study is reported regarding day-ahead electricity load forecasting. For this purpose, three sequence-to-sequence (Seq2seq) deep learning (DL) models are used, namely the multilayer perceptron (MLP), the convolutional neural network (CNN) and the ensemble learning model (ELM), which consists of the weighted combination of the outputs of MLP and CNN models. Also, the study focuses on the development of different forecasting strategies based on DTL, emphasizing the way the datasets are trained and fine-tuned for higher forecasting accuracy. In order to implement the forecasting strategies using deep learning models, load datasets from three Greek islands, Rhodes, Lesvos, and Chios, are used. The main purpose is to apply DTL for day-ahead predictions (1–24 h) for each month of the year for the Chios dataset after training and fine-tuning the models using the datasets of the three islands in various combinations. Four DTL strategies are illustrated. In the first strategy (DTL Case 1), each of the three DL models is trained using only the Lesvos dataset, while fine-tuning is performed on the dataset of Chios island, in order to create day-ahead predictions for the Chios load. In the second strategy (DTL Case 2), data from both Lesvos and Rhodes concurrently are used for the DL model training period, and fine-tuning is performed on the data from Chios. The third DTL strategy (DTL Case 3) involves the training of the DL models using the Lesvos dataset, and the testing period is performed directly on the Chios dataset without fine-tuning. The fourth strategy is a multi-task deep learning (MTDL) approach, which has been extensively studied in recent years. In MTDL, the three DL models are trained simultaneously on all three datasets and the final predictions are made on the unknown part of the dataset of Chios. The results obtained demonstrate that DTL can be applied with high efficiency for day-ahead load forecasting. Specifically, DTL Case 1 and 2 outperformed MTDL in terms of load prediction accuracy. Regarding the DL models, all three exhibit very high prediction accuracy, especially in the two cases with fine-tuning. The ELM excels compared to the single models. More specifically, for conducting day-ahead predictions, it is concluded that the MLP model presents the best monthly forecasts with MAPE values of 6.24% and 6.01% for the first two cases, the CNN model presents the best monthly forecasts with MAPE values of 5.57% and 5.60%, respectively, and the ELM model achieves the best monthly forecasts with MAPE values of 5.29% and 5.31%, respectively, indicating the very high accuracy it can achieve.

The past few years have seen a surge of applying Deep Learning (DL) models for a wide array of tasks such as image classification, object detection, machine translation, etc. While DL models provide an opportunity to solve otherwise intractable tasks, their adoption relies on them being optimized to meet latency and resource requirements. Benchmarking is a key step in this process but has been hampered in part due to the lack of representative and up-to-date benchmarking suites. This is exacerbated by the fast-evolving pace of DL models. This paper proposes DLBricks, a composable benchmark generation design that reduces the effort of developing, maintaining, and running DL benchmarks on CPUs. DLBricks decomposes DL models into a set of unique runnable networks and constructs the original model's performance using the performance of the generated benchmarks. DLBricks leverages two key observations: DL layers are the performance building blocks of DL models and layers are extensively repeated within and across DL models. Since benchmarks are generated automatically and the benchmarking time is minimized, DLBricks can keep up-to-date with the latest proposed models, relieving the pressure of selecting representative DL models. Moreover, DLBricks allows users to represent proprietary models within benchmark suites. We evaluate DLBricks using $50$ MXNet models spanning $5$ DL tasks on $4$ representative CPU systems. We show that DLBricks provides an accurate performance estimate for the DL models and reduces the benchmarking time across systems (e.g. within $95\%$ accuracy and up to $4.4\times$ benchmarking time speedup on Amazon EC2 c5.xlarge).

Deep Learning (DL) has seen rapid adoption in all domains. Since training DL models is expensive, both in terms of time and resources, application workflows that make use of DL increasingly need to operate with a large number of derived learning models, which are obtained through transfer learning and fine-tuning. At scale, thousands of such derived DL models are accessed concurrently by a large number of processes. In this context, an important question is how to design and develop specialized DL model repositories that remain scalable under concurrent access, while addressing key challenges: how to query the DL model architectures for specific patterns? How to load/store a subset of layers/tensors from a DL model? How to efficiently share unmodified layers/tensors between DL models derived from each other through transfer learning? How to maintain provenance and answer ancestry queries? State of art leaves a gap regarding these challenges. To fill this gap, we introduce EvoStore, a distributed DL model repository with scalable data and metadata support to store and access derived DL models efficiently. Large-scale experiments on hundreds of GPUs show significant benefits over state-of-art with respect to I/O and metadata performance, as well as storage space utilization.

COVID-19 is highly infectious and causes acute respiratory disease. Machine learning (ML) and deep learning (DL) models are vital in detecting disease from computerized chest tomography (CT) scans. The DL models outperformed the ML models. For COVID-19 detection from CT scan images, DL models are used as end-to-end models. Thus, the performance of the model is evaluated for the quality of the extracted feature and classification accuracy. There are four contributions included in this work. First, this research is motivated by studying the quality of the extracted feature from the DL by feeding these extracted to an ML model. In other words, we proposed comparing the end-to-end DL model performance against the approach of using DL for feature extraction and ML for the classification of COVID-19 CT scan images. Second, we proposed studying the effect of fusing extracted features from image descriptors, e.g., Scale-Invariant Feature Transform (SIFT), with extracted features from DL models. Third, we proposed a new Convolutional Neural Network (CNN) to be trained from scratch and then compared to the deep transfer learning on the same classification problem. Finally, we studied the performance gap between classic ML models against ensemble learning models. The proposed framework is evaluated using a CT dataset, where the obtained results are evaluated using five different metrics The obtained results revealed that using the proposed CNN model is better than using the well-known DL model for the purpose of feature extraction. Moreover, using a DL model for feature extraction and an ML model for the classification task achieved better results in comparison to using an end-to-end DL model for detecting COVID-19 CT scan images. Of note, the accuracy rate of the former method improved by using ensemble learning models instead of the classic ML models. The proposed method achieved the best accuracy rate of 99.39%.

Accurate rainfall forecasting is crucial in activities such as agricultural planning, water resource management, and flood preparedness, so failure to accurately forecast rainfall might undermine these activities. This study aimed to improve the accuracy of daily rainfall prediction in Hai district, Kilimanjaro region, Tanzania, a region characterised by a bimodal rainfall pattern, by comparing rainfall prediction accuracies of Traditional Machine Learning (TML) and Deep Learning (DL) models. The 25-year daily rainfall dataset for Hai district was first pre-processed by cleaning, normalisation, and splitting into training (first 60%), validation (next 20%), and test (last 20%) sets. TML models (K-Nearest Neighbours (KNN) through its variant KNN Regressor and Support Vector Machine (SVM) through its variant Support Vector Regressor (SVR)) were developed, trained, and their performances compared with developed and trained DL models (Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU)). Each model’s test performance was evaluated using the Root Mean Squared Error (RMSE) metric. The findings revealed that the DL models, LSTM and GRU, achieved test RMSE scores of 6.35 mm and 6.49 mm, respectively, while the TML models, KNN and SVM, achieved test RMSE scores of 6.61 mm and 10.42 mm, respectively. The study demonstrated that DL models are more effective for accurate rainfall forecasting than TML models in regions with complex rainfall patterns. The insights gained will guide the development of more reliable, location-specific weather prediction systems, ultimately supporting informed decision-making in agriculture, environmental management, and flood risk reduction in Tanzania and other regions facing similar climatic challenges

Objective.Functional brain network (FBN) methods are commonly integrated with deep learning (DL) models for EEG analysis. Typically, an FBN is constructed to extract features from EEG data, which are then fed into a DL model for further analysis. Beyond this two-step approach, there is potential to embed FBN construction directly within DL models as a feature extraction module, enabling the models to learn EEG representations end-to-end while incorporating insights from FBNs. However, a critical prerequisite is whether DL models can effectively learn the FBN construction process.Approach.To address this, we propose using DL models to learn FBN matrices derived from EEG data. The ability of DL models to accurately reproduce these matrices would validate their capacity to learn the FBN construction process. This approach is tested on two publicly available EEG datasets, utilizing seven DL models to learn four representative FBN matrices. Model performance is assessed through mean squared error (MSE), Pearson correlation coefficient (Corr), and concordance correlation coefficient (CCC) between predicted and actual matrices.Main results.The results show that DL models achieve low MSE and relatively high Corr and CCC values when learning the Coherence network. Visualizations of predicted and error matrices reveal that while DL models capture the general structure of all four FBNs, certain regions remain difficult to model accurately. Additionally, a pairedt-test comparing global efficiency and nodal degree between predicted and actual networks indicates that most predicted networks significantly differ from the actual networks (p<0.05).Significance.These findings suggest that while DL models can learn the connectivity relationships of certain FBNs, they struggle to capture the intrinsic topological structures. This highlights the irreplaceability of traditional FBN methods in EEG analysis and underscores the need for hybrid strategies that combine FBN methods with DL models for a more comprehensive analysis.

Supervised deep learning (DL) models are becoming ubiquitous in computational chemistry because they can efficiently learn complex input-output relationships and predict chemical properties at a cost significantly lower than methods based on quantum mechanics. The central challenge in many DL applications is the need to invest considerable computational resources in generating large ( N>1×105 ) training sets such that the resulting DL model can be generalized reliably to unseen systems. The lack of better alternatives has encouraged the use of low-cost and relatively inaccurate density-functional theory (DFT) methods to generate training data, leading to DL models that lack accuracy and reliability. In this article, we describe a robust and easily implemented approach based on property-specific atom-centered potentials (ACPs) that resolves this central challenge in DL model development. ACPs are one-electron potentials that are applied in combination with a computationally inexpensive but inaccurate quantum mechanical method (e.g. double-ζ DFT) and fitted against relatively few high-level data ( N≈1×103 – 1×104 ), possibly obtained from the literature. The resulting ACP-corrected methods retain the low cost of the double-ζ DFT approach, while generating high-level-quality data in unseen systems for the specific property for which they were designed. With this approach, we demonstrate that ACPs can be used as an intermediate method between high-level approaches and DL model development, enabling the calculation of large and accurate DL training sets for the chemical property of interest. We demonstrate the effectiveness of the proposed approach by predicting bond dissociation enthalpies, reaction barrier heights, and reaction energies with chemical accuracy at a computational cost lower than the DFT methods routinely used for DL training data set generation.

The bell pepper is a fruit that is often misunderstood as a vegetable due to its usage. The plant which grows bell pepper grows for about three to six feet. These plants do not require intensive care but some necessary steps have to be taken to increase the productivity of the plant. Plant diseases like bacterial spots can heavily affect growth. The consequences of such diseases can be reduced if they’re detected in their early stages. This research aims the comparison of two Deep Learning (DL) models which were developed using VGG-16 architecture and the AlexNet. For training and testing the efficiency of the models, images of the bell pepper plant’s leaf are obtained from Kaggle. This data acquired further undergoes some pre-processing techniques like resizing and rescaling to reach the desired format. The cleaned data is then split into three divisions. One of them is used to train both DL models and the second one is used to validate the models. The last part is used to test the accuracy and loss of the two models. In the end, the accuracy and loss of both the models are compared and the results are tabulated.

Soil temperature (ST) plays an important role in agriculture and other fields, and has a close relationship with plant growth and development. Therefore, accurate ST prediction methods are widely needed. Deep learning (DL) models have been widely applied for ST prediction. However, the traditional DL models may fail to capture the spatiotemporal relationship due to its complex dependency under different related hydrologic variables. Hence, the DL models with Ensemble Empirical Mode Decomposition (EEMD) are proposed in this study. The proposed models can capture more complex spatiotemporal relationship after decomposing the ST into different intrinsic mode functions. Therefore, the performance of models is further improved. The results show that the performance of DL models with EEMD are better than that of corresponding DL models without EEMD. Moreover, EEMD-Conv3d has the best performance among all the experimental models. It has the highest R2 ranging from 0.9826 to 0.9893, the lowest RMSE ranging from 1.3096 to 1.6497 and the lowest MAE ranging from 0.9656 to 1.2056 in predicting ST at the lead time from one to five days. In addition, the lines between predicted ST and observed ST are closer to the ideal line (y = x) than other DL models. The results show that our EEMD-Conv3D can better capture spatiotemporal correlation and is an applicable method for predicting spatiotemporal ST.

We developed a multimodal ultrasound (US) deep learning (DL) fusion model to automatically classify early fibrosis in patients with chronic kidney disease (CKD). This prospective study included patients with CKD who underwent continuous gray-scale US, superb microvascular imaging, and strain elastography from May to November 2022. According to the pathological tubular atrophy and interstitial fibrosis score, patients were divided into minimal and mild groups (affected area ≤10% and 11 − 25% of the total cortical volume, respectively). The dataset was divided into training (70%) and test (30%) sets. A DL model combining the features of the three US modes was developed to predict early fibrosis in patients with CKD. We compared these findings with the area under the receiver operating characteristic curve (AUC) of the clinical model by analyzing the receiver operating characteristic curve in the test set. The AUC of single-mode DL based on gray-scale US, superb microvascular imaging, and strain elastography was 0.682, 0.745, and 0.648, respectively, while that of the multimodal US DL model was 0.86. The accuracy, specificity, and sensitivity of the multimodal US DL model were 0.779, 0.767, and 0.796, respectively, and the negative and positive predictive values were 0.842 and 0.706, respectively. The AUC of the multimodal US DL model was significantly better than that of the single-mode DL and clinical models. The DL algorithm developed using multimodal US images can effectively predict early fibrosis in patients with CKD with significantly greater accuracy than single-mode DL or clinical models.

We aimed at comparing intratumoral and peritumoral deep learning, radiomics, and fusion models in predicting KRAS mutations in rectal cancer using endorectal ultrasound imaging. This study included 304 patients with rectal cancer from Fujian Medical University Union Hospital. The patients were randomly divided into a training group (213 patients) and a test group (91 patients) at a 7:3 ratio. Radiomics and deep learning models were established using primary tumor and peritumoral images. In the optimally performing regions-of-interest, two fusion strategies, a feature-based and a decision-based model, were employed to build the fusion models. The Shapley additive explanation (SHAP) method was used to evaluate the significance of features in the optimal radiomics, deep learning, and fusion models. The performance of each model was assessed using the area under the receiver operating characteristic curve (AUC) and decision curve analysis (DCA). In the test cohort, both the radiomics and deep learning models exhibited optimal performance with a 10-pixel patch extension, yielding AUC values of 0.824 and 0.856, respectively. The feature-based DLRexpand10_FB model attained the highest AUC (0.896) across all study sets. In addition, the DLRexpand10_FB model demonstrated excellent sensitivity, specificity, and DCA. SHAP analysis underscored the deep learning feature (DL_1) as the most significant factor in the hybrid model. The feature-based fusion model DLRexpand10_FB can be employed to predict KRAS gene mutations based on pretreatment endorectal ultrasound images of rectal cancer. The integration of peritumoral regions enhanced the predictive performance of both the radiomics and deep learning models.