Data Transfer Learning Research Articles

Assessment of artificial intelligence to detect gasoline in fire debris using HS-SPME-GC/MS and transfer learning.

Due to the complex nature of the chemical compositions of ignitable liquids (IL) and the interferences from fire debris matrices, interpreting chromatographic data poses challenges to analysts. In this work, artificial intelligence (AI) was developed by transfer learning in a convolutional neural network (CNN), GoogLeNet. The image classification AI was fine-tuned to create intelligent classification systems to discriminate samples containing gasoline residues from burned substrates. All ground truth samples were analyzed by headspace solid-phase microextraction (HS-SPME) coupled with a gas chromatograph and mass spectrometer (GC/MS). The HS-SPME-GC/MS data were transformed into three types of image presentations, that is, heatmaps, extracted ion heatmaps, and total ion chromatograms. The abundance and mass-to-charge ratios of each scan were converted into image patterns that are characteristic of the chemical profiles of gasoline. The transfer learning data were labeled as "gasoline present" and "gasoline absent" classes. The assessment results demonstrated that all AI models achieved 100 ± 0% accuracy in identifying neat gasoline. When the models were assessed using the spiked samples, the AI model developed using the extracted ion heatmap obtained the highest accuracy rate (95.9 ± 0.4%), which was greater than those obtained by other machine learning models, ranging from 17.3 ± 0.7% to 78.7 ± 0.7%. The proposed work demonstrated that the heatmaps created from GC/MS data can represent chemical features from the samples. Additionally, the pretrained CNN models are readily available in the transfer learning workflow to develop AI for GC/MS data interpretation in fire debris analysis.

Monitoring the misalignment of machine tools with autoencoders after they are trained with transfer learning data

CNC machines have revolutionized manufacturing by enabling high-quality and high-productivity production. Monitoring the condition of these machines during production would reduce maintenance cost and avoid manufacturing defective parts. Misalignment of the linear tables in CNCs can directly affect the quality of the manufactured parts, and the components of the linear tables wear out over time due to the heavy and fluctuating loads. To address these challenges, an intelligent monitoring system was developed to identify normal operation and misalignments. Since damaging a CNC machine for data collection is too expensive, transfer learning was used in two steps. First, a specially designed experimental feed axis test platform (FATP) was used to sample the current signal at normal and five levels of left-side misalignment conditions ranging from 0.05 to 0.25 mm. Four different algorithm combinations were trained to detect misalignments. These combinations included a 1D convolution neural network (CNN) and autoencoder (AE) combination, a temporal convolutional network (TCN) and AE combination, a long short-term memory neural network (LSTM) and AE combination, and a CNN, LSTM, and AE combination. At the second step, Wasserstein deep convolutional generative adversarial network (W-DCGAN) was used to generate data by integrating the observed characteristics of the FATP at different misalignment levels and collected limited data from the actual CNC machines. To evaluate the similarity and limited diversity of generated and real signals, t-distributed stochastic neighbor embedding (T-SNE) method was used. The hyperparameters of the model were optimized by random and grid search. The CNN, LSTM, and AE combination demonstrated the best performance, which provides a practical way to detect misalignments without stopping production or cluttering the work area with sensors. The proposed intelligent monitoring system can detect misalignments of the linear tables of CNCs, thus enhancing the quality of manufactured parts and reducing production costs.

Anatomically guided self-adapting deep neural network for clinically significant prostate cancer detection on bi-parametric MRI: a multi-center study

ObjectiveTo evaluate the effectiveness of a self-adapting deep network, trained on large-scale bi-parametric MRI data, in detecting clinically significant prostate cancer (csPCa) in external multi-center data from men of diverse demographics; to investigate the advantages of transfer learning.MethodsWe used two samples: (i) Publicly available multi-center and multi-vendor Prostate Imaging: Cancer AI (PI-CAI) training data, consisting of 1500 bi-parametric MRI scans, along with its unseen validation and testing samples; (ii) In-house multi-center testing and transfer learning data, comprising 1036 and 200 bi-parametric MRI scans. We trained a self-adapting 3D nnU-Net model using probabilistic prostate masks on the PI-CAI data and evaluated its performance on the hidden validation and testing samples and the in-house data with and without transfer learning. We used the area under the receiver operating characteristic (AUROC) curve to evaluate patient-level performance in detecting csPCa.ResultsThe PI-CAI training data had 425 scans with csPCa, while the in-house testing and fine-tuning data had 288 and 50 scans with csPCa, respectively. The nnU-Net model achieved an AUROC of 0.888 and 0.889 on the hidden validation and testing data. The model performed with an AUROC of 0.886 on the in-house testing data, with a slight decrease in performance to 0.870 using transfer learning.ConclusionsThe state-of-the-art deep learning method using prostate masks trained on large-scale bi-parametric MRI data provides high performance in detecting csPCa in internal and external testing data with different characteristics, demonstrating the robustness and generalizability of deep learning within and across datasets.Clinical relevance statementA self-adapting deep network, utilizing prostate masks and trained on large-scale bi-parametric MRI data, is effective in accurately detecting clinically significant prostate cancer across diverse datasets, highlighting the potential of deep learning methods for improving prostate cancer detection in clinical practice.Graphical

Exploring nonlinear strengthening in polycrystalline metallic materials by machine learning methods and heterostructure design

To improve the strength and plasticity of structural materials, researchers often introduce various strengthening mechanisms such as second-phase strengthening, dislocation strengthening, and back stress strengthening (HDI). Due to the interaction of multiple mechanisms, the linear superposition relationship has a poor fitting effect and is only used for rough calculations of the strengthening mechanisms. In this study, the transfer learning data was used to optimize the deep learning network structure (Re-CNN) based on the residual algorithm, and the yield strength prediction physical neural informed model (PNIM) of polycrystalline metallic materials was established. To promote the industrial application of the heterostructure design method, a medium carbon steel heterostructure design strategy based on the existing equipment of the factory was proposed. Medium-carbon heterostructure materials (MHSM) with mixed strengthening mechanisms were successfully prepared. MHSM exhibits excellent comprehensive mechanical properties. When a linear relationship is used to describe the MHSM yield strength, there is a large error, while Re-CNN shows satisfactory prediction accuracy. The linear relationship is incompatible with homogeneous structure materials and heterogeneous structure materials, and its universality is lower than that of nonlinear Re-CNN. Re-CNN shows high cross-scale prediction ability and can be compatible with homogeneous microstructures and heterogeneous microstructures. Using the heterogeneity evolution characteristics of MHSM, the key factors deviating from the linear relationship were revealed. The overestimation and underestimation of the linear relation are demonstrated by Taylor factor and TEM analysis to be caused by the multiscale properties of ferrite, the behavior of the second phase particles, and the interaction of various mechanisms. This study provides a new idea for the cross-scale calculation of the mechanical properties of polycrystalline metallic materials.

Deep Unsupervised Representation Learning for Feature-Informed EEG Domain Extraction.

In electroencephalography (EEG) classification paradigms, data from a target subject is often difficult to obtain, leading to difficulties in training a robust deep learning network. Transfer learning and their variations are effective tools in improving such models suffering from lack of data. However, many of the proposed variations and deep models often rely on a single assumed distribution to represent the latent features which may not scale well due to inter- and intra-subject variations in signals. This leads to significant instability in individual subject decoding performances. The presence of non-trivial domain differences between different sets of training or transfer learning data causes poorer model generalization towards the target subject. However, the detection of these domain differences is often difficult to perform due to the ill-defined nature of the EEG domain features. This study proposes a novel inference model, the Joint Embedding Variational Autoencoder, that offers conditionally tighter approximation of the estimated spatiotemporal feature distribution through the use of jointly optimised variational autoencoders to achieve optimizable data dependent inputs as an additional variable for improved overall model optimisation and scaling without sacrificing model tightness. To learn the variational bound, we show that maximising the marginal log-likelihood of only the second embedding section is required to achieve conditionally tighter lower bounds. Furthermore, we show that this model provides state-of-the-art EEG data reconstruction and deep feature extraction. The extracted domains of the EEG signals across each subject displays the rationale as to why there exists disparity between subjects' adaptation efficacy.

A Novel Approach to the Job Shop Scheduling Problem Based on the Deep Q-Network in a Cooperative Multi-Access Edge Computing Ecosystem.

In this study, based on multi-access edge computing (MEC), we provided the possibility of cooperating manufacturing processes. We tried to solve the job shop scheduling problem by applying DQN (deep Q-network), a reinforcement learning model, to this method. Here, to alleviate the overload of computing resources, an efficient DQN was used for the experiments using transfer learning data. Additionally, we conducted scheduling studies in the edge computing ecosystem of our manufacturing processes without the help of cloud centers. Cloud computing, an environment in which scheduling processing is performed, has issues sensitive to the manufacturing process in general, such as security issues and communication delay time, and research is being conducted in various fields, such as the introduction of an edge computing system that can replace them. We proposed a method of independently performing scheduling at the edge of the network through cooperative scheduling between edge devices within a multi-access edge computing structure. The proposed framework was evaluated, analyzed, and compared with existing frameworks in terms of providing solutions and services.

A Generative Model to Synthesize EEG Data for Epileptic Seizure Prediction.

Objective: Scarcity of good quality electroencephalography (EEG) data is one of the roadblocks for accurate seizure prediction. This work proposes a deep convolutional generative adversarial network (DCGAN) to generate synthetic EEG data. Another objective of our study is to use transfer-learning (TL) for evaluating the performance of four well-known deep-learning (DL) models to predict epileptic seizure. Methods: We proposed an algorithm that generate synthetic data using DCGAN trained on real EEG data in a patient-specific manner. We validate quality of generated data using one-class SVM and a new proposal namely convolutional epileptic seizure predictor (CESP). We evaluate performance of VGG16, VGG19, ResNet50, and Inceptionv3 trained on augmented data using TL with average time of 10 min between true prediction and seizure onset samples. Results: The CESP model achieves sensitivity of 78.11% and 88.21%, and false prediction rate of 0.27/h and 0.14/h for training on synthesized and testing on real Epilepsyecosystem and CHB-MIT datasets, respectively. Using TL and augmented data, Inceptionv3 achieved highest accuracy with sensitivity of 90.03% and 0.03 FPR/h. With the proposed data augmentation method prediction results of CESP model and Inceptionv3 increased by 4-5% as compared to state-of-the-art augmentation techniques. Conclusion: The performance of CESP shows that synthetic data acquired association between features and labels very well and by using the augmented data CESP predicted better than chance level for both datasets. Significance: The proposed DCGAN can be used to generate synthetic data to increase the prediction performance and to overcome good quality data scarcity issue.

DeepVesselNet: Vessel Segmentation, Centerline Prediction, and Bifurcation Detection in 3-D Angiographic Volumes.

We present DeepVesselNet, an architecture tailored to the challenges faced when extracting vessel trees and networks and corresponding features in 3-D angiographic volumes using deep learning. We discuss the problems of low execution speed and high memory requirements associated with full 3-D networks, high-class imbalance arising from the low percentage (<3%) of vessel voxels, and unavailability of accurately annotated 3-D training data—and offer solutions as the building blocks of DeepVesselNet. First, we formulate 2-D orthogonal cross-hair filters which make use of 3-D context information at a reduced computational burden. Second, we introduce a class balancing cross-entropy loss function with false-positive rate correction to handle the high-class imbalance and high false positive rate problems associated with existing loss functions. Finally, we generate a synthetic dataset using a computational angiogenesis model capable of simulating vascular tree growth under physiological constraints on local network structure and topology and use these data for transfer learning. We demonstrate the performance on a range of angiographic volumes at different spatial scales including clinical MRA data of the human brain, as well as CTA microscopy scans of the rat brain. Our results show that cross-hair filters achieve over 23% improvement in speed, lower memory footprint, lower network complexity which prevents overfitting and comparable accuracy that does not differ from full 3-D filters. Our class balancing metric is crucial for training the network, and transfer learning with synthetic data is an efficient, robust, and very generalizable approach leading to a network that excels in a variety of angiography segmentation tasks. We observe that sub-sampling and max pooling layers may lead to a drop in performance in tasks that involve voxel-sized structures. To this end, the DeepVesselNet architecture does not use any form of sub-sampling layer and works well for vessel segmentation, centerline prediction, and bifurcation detection. We make our synthetic training data publicly available, fostering future research, and serving as one of the first public datasets for brain vessel tree segmentation and analysis.

Weighted Transfer Learning of Dynamic Time Warped Data for Motor Imagery based Brain Computer Interfaces.

A large amount of calibration data is typically needed to train an electroencephalogram (EEG)-based brain-computer interfaces (BCI) due to the non-stationary nature of EEG data. This paper proposes a novel weighted transfer learning algorithm using a dynamic time warping (DTW) based alignment method to alleviate this need by using data from other subjects. DTW-based alignment is first applied to reduce the temporal variations between a specific subject data and the transfer learning data from other subjects. Next, similarity is measured using Kullback Leibler divergence (KL) between the DTW aligned data and the specific subject data. The other subjects' data are then weighted based on their KL similarity to each trials of the specific subject data. This weighted data from other subjects are then used to train the BCI model of the specific subject. An experiment was performed on publicly available BCI Competition IV dataset 2a. The proposed algorithm yielded an average improvement of 9% compared to a subject-specific BCI model trained with 4 trials, and the results yielded an average improvement of 10% compared to naive transfer learning. Overall, the proposed DTW-aligned KL weighted transfer learning algorithm show promise to alleviate the need of large amount of calibration data by using only a short calibration data.

Matching Component Analysis for Transfer Learning

We introduce a new Procrustes-type method called matching component analysis to isolate components in data for transfer learning. Our theoretical results describe the sample complexity of this method, and we demonstrate through numerical experiments that our approach is indeed well suited for transfer learning.