Target Task Research Articles

Transferability prediction among classification and regression tasks using optimal transport

Transfer learning is a method for improving generalization performance by training a model for a different task first and then additionally training the pre-learned weights for the target task. However, transferability—the ease with which source task can be effectively transferred to which target task—is often unknown. Existing works proposed methods of measuring the transferability between classification tasks using images and discrete labels, but it cannot be applied to regression tasks. In this work, we investigate transferability among classification and regression tasks, and propose a method for predicting transferability by extending the optimal transport theory. Our transferability prediction model also can be applied to subjective tasks (e.g., aesthetics and memorability), which are usually regression tasks. We show that the appropriate source (pre-training) tasks can be predicted for the chosen target task without conducting actual pre-training and transferring trials. Experimental results demonstrated high prediction accuracy (correlation coefficient of ρ=0.791\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\rho =0.791$$\\end{document}) and a speed improvement of approximately 300 times compared with the above-mentioned greedy approach.

Optimizing Image Processing with CNNs through Transfer Learning: Survey

The field of image processing has been revolutionized by Convolutional Neural Networks (CNNs), which exhibit exceptional capability in feature extraction and accurate image classification. However, training CNNs requires large volumes of annotated data and significant computational resources. Considering these challenges, transfer learning has emerged as a promising approach to reducing the dependence on labeled data and computational resources. Transfer learning involves utilizing knowledge gained from a source task to improve the training process for a target task. This technique has demonstrated considerable benefits; however, it also possesses certain limitations. Consequently, this survey explores the advantages and constraints of transfer learning and the various factors that influence its effectiveness in optimizing image processing using CNNs. Additionally, the survey investigates the most recent advancements and research in the field of transfer learning specifically for image processing with CNNs. In summary, this comprehensive analysis highlights the significance of transfer learning in the context of optimizing image processing with CNNs, providing unique insights into this rapidly evolving domain.

Deep reinforcement learning for cooperative robots based on adaptive sentiment feedback

Human–robot cooperative tasks have gained importance with the emergence of robotics and artificial intelligence technology. In interactive reinforcement learning techniques, robots learn target tasks by receiving feedback from an experienced human trainer. However, most interactive reinforcement learning studies require a separate process to integrate the trainer’s feedback into the training dataset, making it challenging for robots to learn new tasks from humans in real-time. Furthermore, the types of feedback sentences that trainers can use are limited in previous research. To address these limitations, this paper proposes a robot teaching strategy that uses deep RL via human–robot interaction to learn table balancing tasks interactively. The proposed system employs Deep Q-Network with real-time sentiment feedback delivered through the trainer’s speech to learn cooperative tasks. We designed a novel reward function that incorporates sentiment feedback from human speech in real-time during the learning process. The paper presents an improved reward shaping technique based on subdivided feedback levels and shrinking feedback. This function serves as a guide for the robot to engage in natural interactions with humans and enables it to learn the tasks effectively. Experimental results demonstrate that the proposed interactive deep reinforcement learning model achieved a high success rate of up to 99.06%, outperforming the model without sentiment feedback.

Attention-based multimodal fusion with contrast for robust clinical prediction in the face of missing modalities

Objective:With the increasing amount and growing variety of healthcare data, multimodal machine learning supporting integrated modeling of structured and unstructured data is an increasingly important tool for clinical machine learning tasks. However, it is non-trivial to manage the differences in dimensionality, volume, and temporal characteristics of data modalities in the context of a shared target task. Furthermore, patients can have substantial variations in the availability of data, while existing multimodal modeling methods typically assume data completeness and lack a mechanism to handle missing modalities. Methods:We propose a Transformer-based fusion model with modality-specific tokens that summarize the corresponding modalities to achieve effective cross-modal interaction accommodating missing modalities in the clinical context. The model is further refined by inter-modal, inter-sample contrastive learning to improve the representations for better predictive performance. We denote the model as Attention-based cRoss-MOdal fUsion with contRast (ARMOUR). We evaluate ARMOUR using two input modalities (structured measurements and unstructured text), six clinical prediction tasks, and two evaluation regimes, either including or excluding samples with missing modalities. Results:Our model shows improved performances over unimodal or multimodal baselines in both evaluation regimes, including or excluding patients with missing modalities in the input. The contrastive learning improves the representation power and is shown to be essential for better results. The simple setup of modality-specific tokens enables ARMOUR to handle patients with missing modalities and allows comparison with existing unimodal benchmark results. Conclusion:We propose a multimodal model for robust clinical prediction to achieve improved performance while accommodating patients with missing modalities. This work could inspire future research to study the effective incorporation of multiple, more complex modalities of clinical data into a single model.

Non-destructive detection method for wheat freshness degree based on delayed luminescence

Accurate identification to freshness degree is one of the preconditions for secure storage of wheat. Up to now, the detection on wheat freshness mainly depends on biochemical methods. However, biochemical methods are always involved in a series of problems, such as long pretreatment processes to wheat samples, poor repeatability, destructive detection and so on, which hardly meet the detection requirements of scientific and intelligent grain storage. Therefore, finding out a simple, accurate, and non-destructive detection method for wheat freshness degree is of necessity. For this reason, we propose a novel wheat freshness degree detection model based on the delayed luminescence (DL) signals of wheat samples in this paper. Furthermore, a bidirectional LSTM network based on Walsh coding (Walsh-Bi-LSTM) is introduced in order to make the detection model have the error-correcting performance by reasonably splitting the multi-classification target task into several binary classification targets. Shown by the experimental results, the classification accuracy rates of the detection model established in this paper achieve 90% and 94% on the test sets of storage wheat and artificial aging seed wheat samples, improving 9% and 12% compared with conventional Bi-LSTM network model respectively, which validates that the proposed model can accurately detect wheat freshness degree.

Influence of target location, task complexity, and gender on children's use of their preferred foot.

The goal of the present study was to compare the choice of foot of right-footed children with varying strengths of foot preference when performing two tasks of different levels of complexity at three spatial locations. 30 right-footed children were tested. The results showed that the general tendency to use one's preferred foot to interact with an object is more or less pronounced depending on the object's location and the complexity of the task. The children used their non-preferred left foot more often during the simple task and when the object was presented to the left. Our findings also revealed interactions between gender, age, and spatial location. At five, girls used globally less their left foot than boys. In addition, girls used their preferred right foot more frequently than boys when the object was presented to the left. Based on these results, it can be concluded that foot selection depends on foot preference, task demand, environmental context, and biological factors associated with motor dominance.

A novel momentum prototypical neural network to cross-domain fault diagnosis for rotating machinery subject to cold-start

Cross-domain rotating machinery fault diagnosis has achieved great success recently with the development of deep transfer learning. However, conventional deep transfer learning methods encounter a severe decline in prediction accuracy when fault samples are limited. Moreover, conventional deep transfer learning methods require additional parameter tuning rather than cold-start when applied to the target tasks, hampering their implementation in practical fault diagnosis applications. In this paper, a novel method, named momentum prototypical neural network (MoProNet), is proposed for cross-domain few-shot rotating machinery fault diagnosis. The MoProNet progressively updates the support encoder to address the prototype oscillation problem and enable the model to apply limited source domain samples to predict target domain faults with cold-start. The performance of the proposed MoProNet is tested on a bearing dataset and a hardware-in-the-loop high-speed train simulation platform, respectively, with over forty cross-domain few-shot fault diagnosis tasks. The experimental results demonstrate that the proposed MoProNet achieves satisfactory results and outperforms the other comparable methods in the same cross-domain few-shot scenarios with the simple AlexNet backbone.

Content preserving image translation with texture co-occurrence and spatial self-similarity for texture debiasing and domain adaptation

Models trained on datasets with texture bias usually perform poorly on out-of-distribution samples since biased representations are embedded into the model. Recently, various image translation and debiasing methods have attempted to disentangle texture biased representations for downstream tasks, but accurately discarding biased features without altering other relevant information is still challenging. In this paper, we propose a novel framework that leverages image translation to generate additional training images using the content of a source image and the texture of a target image with a different bias property to explicitly mitigate texture bias when training a model on a target task. Our model ensures texture similarity between the target and generated images via a texture co-occurrence loss while preserving content details from source images with a spatial self-similarity loss. Both the generated and original training images are combined to train improved classification or segmentation models robust to inconsistent texture bias. Evaluation on five classification- and two segmentation-datasets with known texture biases demonstrates the utility of our method, and reports significant improvements over recent state-of-the-art methods in all cases.

Enhancing the Accuracy of an Image Classification Model Using Cross-Modality Transfer Learning

Applying deep learning (DL) algorithms for image classification tasks becomes more challenging with insufficient training data. Transfer learning (TL) has been proposed to address these problems. In theory, TL requires only a small amount of knowledge to be transferred to the target task, but traditional transfer learning often requires the presence of the same or similar features in the source and target domains. Cross-modality transfer learning (CMTL) solves this problem by learning knowledge in a source domain completely different from the target domain, often using a source domain with a large amount of data, which helps the model learn more features. Most existing research on CMTL has focused on image-to-image transfer. In this paper, the CMTL problem is formulated from the text domain to the image domain. Our study started by training two separately pre-trained models in the text and image domains to obtain the network structure. The knowledge of the two pre-trained models was transferred via CMTL to obtain a new hybrid model (combining the BERT and BEiT models). Next, GridSearchCV and 5-fold cross-validation were used to identify the most suitable combination of hyperparameters (batch size and learning rate) and optimizers (SGDM and ADAM) for our model. To evaluate their impact, 48 two-tuple hyperparameters and two well-known optimizers were used. The performance evaluation metrics were validation accuracy, F1-score, precision, and recall. The ablation study confirms that the hybrid model enhanced accuracy by 12.8% compared with the original BEiT model. In addition, the results show that these two hyperparameters can significantly impact model performance.

Study on Human Capital Structure Assessment and Relative Poverty Governance

China has completed the target task of fully eradicating poverty among the absolute rural poor under the current standards on schedule, and the milestone victory in poverty alleviation means that the focus of poverty alleviation will shift from helping absolute poverty in the past to targeting relative poverty. Human capital is always a fundamental intrinsic factor in alleviating the problem of relative poverty. This paper completes action on the usual policy of shifting the focus of China's poverty alleviation strategy to comparative poverty management, and explores how the optimization of human capital structure affects the alleviation of comparative poverty, and what is its intrinsic influence logic? Using spatial vector theory and coupled synergy model, we construct a model to assess the level of human capital structure from the perspectives of both advanced human capital and high-low skill complementarity, and conduct regression analysis to examine the total effect of human capital structure optimization on relative poverty, and use methods such as mediating effect model to explore its intrinsic influence mechanism in depth, and conduct robustness test to conclude that human capital structure optimization has a positive effect on the alleviation of relative poverty. The results show that the optimization of human capital structure has a positive effect on the alleviation of virtual poverty. This study provides policy suggestions and paths to reduce relative poverty through optimizing human capital structure in the post-poverty era.

Interpretability Study of Transfer Learning for Emotion Classification

Transfer learning is a powerful technique for improving the performance of deep neural networks on a target task with limited labeled data. However, it is not always clear why pre-trained models on unrelated tasks, such as ImageNet, can be effective for emotion recognition tasks. This paper evaluates the performance of transfer learning for facial expression recognition tasks and investigates the influence of network depth on performance using three different typical pre-trained models: VGG16, MobileNet, and ResNet50. The layers of the pre-trained models are frozen, and their output layer is substituted with a new output layer that corresponds to the seven-class emotion classes in the target dataset. The study demonstrates that transfer learning with pre-trained models, even though with limited labelled dataset regarding facial expressions, can be an effective approach for improving the performance of deep neural networks on emotion recognition tasks. The findings provide insights into the interpretability of transfer learning for facial expression recognition tasks and highlight the importance of fine-tuning pre-trained models for optimal performance. Additionally, the study also offers ideas for future model design in facial expression recognition tasks.

Evolutionary Multitasking for Optimization Based on Generative Strategies

Evolutionary multitasking (EMT) is one of the emerging topics in evolutionary computation. EMT can solve multiple related optimization tasks simultaneously and enhance the optimization of each task via knowledge sharing among tasks. Many EMT algorithms have been proposed and achieved success in various problems, yet EMT for multi-objective optimization remains a big challenge. The existing multi-objective EMT algorithms tend to suffer from slow convergence and difficulty in generating high-quality knowledge. To alleviate these issues, this paper proposes a new EMT algorithm namely EMT-GS for multi-objective optimization based on two generative strategies. Particularly, generative adversarial networks and inertial differential evolution are introduced to generate transferable knowledge and offspring, respectively. A generative adversarial network is trained periodically for each source-target task pair, based on which helpful knowledge is generated from the source task and transferred to boost the solving of the target task. To accelerate the population convergence, the inertial differential evolution strategy is put forward to generate offspring in a promising direction according to the individuals from the previous generation and the transferred knowledge. The performance of EMT-GS is validated on three multitasking multi-objective benchmark problems. The experimental results highlight the excellent competitiveness of EMT-GS compared to other state-of-the-art multi-objective EMT algorithms.

ANAS: Asymptotic NAS for large-scale proxyless search and multi-task transfer learning

Neural Architecture Search (NAS) is an emerging solution to design a lightweight network for researchers to obtain a trade-off between accuracy and speed, releasing researchers from tedious repeated trials. However, the main shortcoming of NAS is its high and unstable memory consumption of the search work, especially for large-scale tasks. In this study, the proposed Asymptotic Neural Architecture Search network (ANAS) achieved a proxyless search for large-scale tasks with economic and stable memory consumption. Instead of proxy search like other NAS algorithms, ANAS achieved the large-scale proxyless search that directly learns deep neural network architecture for target task. ANAS reduced the peak value of memory consumption by an asymptotic method, and kept the memory consumption stable by the linkage change of a series of key indexes. A pruning operation and efficient candidate operations decreased the total memory consumption. Finally, ANAS achieved a good trade-off between accuracy and speed for classification tasks on CIFAR-10, CIFAR-100, and ImageNet datasets. Besides, except for the classification task, it achieved excellent multi-task transfer learning ability for implementing the segmentation task on CamVid and Cityscapes. ANAS reached 22.8% test errs with 5 M parameter on ImageNet, and 72.9 mIoU (mean Intersection over Union) with 119.9 FPS (Frames Per Second) on Cityscapes dataset.

Revisiting the transferability of adversarial examples via source-agnostic adversarial feature inducing method

Though deep neural networks (DNNs) have revealed their extraordinary performance in the fields of computer vision, it is evident that the vulnerability of DNNs to adversarial attacks with crafted human-imperceptible perturbations. Most existing adversarial attacks draw their attention to invading target deep task models by enhancing input-diagnostic features via image rotation, warp, or transformation to improve adversarial transferability. Such manners pay close concentration to operation on original inputs regardless of the properties from different source information. Research has inspired us to consider utilizing source-agnostic information and integrating generated features with raw inputs to enrich adversarial properties. For such needs, we propose a simple and flexible adversarial attack method with source-agnostic Feature Inducing Method (FIM) for improving the transferability of adversarial examples (AEs). FIM first focuses on generating perturbed features by imitating diverse patterns from multi-domain sources. Instead of exploiting the original inputs’ diversity, such proposed work gains the various properties by random feature imitation referring to different source distributions. By optimizing the generated features with norm bounds, FIM then integrates original inputs with imitative features. Such manner can diverse row positive class-general features, which reduce the capability of class-specific patterns on cross-model transferability. Based on the crafted property, FIM employs the adaptive gradient-based strategy on such information to generate perturbations, which helps to decrease probability dropping into local optimal when searching for the decision boundary of source and target models. We conduct detailed experiments to evaluate the performance of our proposed approach with existing baselines on three public datasets. The experimental results reveal the better performance of the proposed works on fooling source and target task models leading to a considerable margin in most adversarial scenarios. We further investigate adversarial attacks on adversarial defense models (with adversarial training and trades). Such a proposed attack strategy achieves better attack quality by a margin over 3.00% on CIFAR10 and reduces the robust accuracy of adversarially trained models by a large margin near 9.00% on MNIST. Furthermore, we exploit the performance of the proposed attack strategy applied to feature-level adversarial domains and conduct evaluations to demonstrate its adversarial feasibility in integrating with various attack mechanisms, which gains better adversarial effectiveness over 20.00% than the base attacks on studied deep task models.

Formula omitted]: Tracing and locating the key elements in sentiment analysis

We study sentiment analysis task where the outcomes are mainly contributed by a few key elements of the inputs. Motivated by the two-streams hypothesis, we explore processing input items and their weights separately by developing a neural architecture, named TraceNet, to address this type of task. It not only learns discriminative representations for the target task via its encoders, but also traces key elements at the same time via its locators. In TraceNet, both encoders and locators are organized in a layer-wise manner, and a smoothness regularization is employed between adjacent encoder-locator combinations. Moreover, a sparsity constraint is enforced on locators for tracing purposes and items are proactively masked according to the item weights output by locators. A major advantage of TraceNet is that the outcomes are easier to understand, as it identifies the key components responsible for the outcomes, making them easier to understand. The experimental results demonstrate its effectiveness in sentiment classification. Furthermore, we present case studies to showcase the interpretability of the model and conduct comprehensive analyses to highlight the impacts of each component.

Multifaceted Relation-aware Meta-learning with Dual Customization for User Cold-start Recommendation

User cold-start scenarios pose great challenges to recommendation systems in accurately capturing user preferences with sparse interaction records. Besides incorporating auxiliary information to enrich user/item representations, recent studies under the schema of meta-learning focus on quickly adapting personalized recommendation models based on cold-start users’ scarce interactions. The majority of meta-learning based recommendation methods follow a bi-level optimization paradigm and learn globally shared initialization across all cold-start recommendation tasks. In addition, to further facilitate the ability of fast adaptation, existing methods have made efforts to tailor task-specific prior knowledge by identifying the individual characteristics of each task. However, we argue that multi-view commonalities between existing users and cold-start users are also essential for precisely distinguishing new tasks, but not comprehensively modeled in previous studies. In this article, we propose a multifaceted relation-aware meta-learning approach namely MeCM for user cold-start recommendation, which enhances task-adaptive initialization customization by extracting multiple views of task relevance. We design a dual customization framework consisting of two successive phases including cluster-level customization and task-level customization. Specifically, MeCM first extracts multifaceted semantic relations between tasks and refines task commonalities into task clusters maintained with memory networks (MNs). Globally learned fast weights corresponding to task clusters are queried to perform cluster-level customization. Then task-level customization is triggered based on contextual information of the target task via interaction-wise encoding. Extensive experiments on real-world datasets demonstrate the superior performance of our model over state-of-the-art meta-learning-based recommendation methods.

Subtask-masked curriculum learning for reinforcement learning with application to UAV maneuver decision-making

Unmanned Aerial Vehicle (UAV) maneuver strategy learning remains a challenge when using Reinforcement Learning (RL) in this sparse reward task. In this paper, we propose Subtask-Masked curriculum learning for RL (SubMas-RL), an efficient RL paradigm that implements curriculum learning and knowledge transfer for UAV maneuver scenarios involving multiple missiles. First, this study introduces a novel concept known as subtask mask to create source tasks from a target task by masking partial subtasks. Then, a subtask-masked curriculum generation method is proposed to generate a sequenced curriculum by alternately conducting task generation and task sequencing. To establish efficient knowledge transfer and avoid negative transfer, this paper employs two transfer techniques, policy distillation and policy reuse, along with an explicit transfer condition that masks irrelevant knowledge. Experimental results demonstrate that our method achieves a 94.8% success rate in the UAV maneuver scenario, where the direct use of reinforcement learning always fails. The proposed RL framework SubMas-RL is expected to learn an effective policy in complex tasks with sparse rewards.

Automaton-Guided Curriculum Generation for Reinforcement Learning Agents

Despite advances in Reinforcement Learning, many sequential decision making tasks remain prohibitively expensive and impractical to learn. Recently, approaches that automatically generate reward functions from logical task specifications have been proposed to mitigate this issue; however, they scale poorly on long-horizon tasks (i.e., tasks where the agent needs to perform a series of correct actions to reach the goal state, considering future transitions while choosing an action). Employing a curriculum (a sequence of increasingly complex tasks) further improves the learning speed of the agent by sequencing intermediate tasks suited to the learning capacity of the agent. However, generating curricula from the logical specification still remains an unsolved problem. To this end, we propose AGCL, Automaton-guided Curriculum Learning, a novel method for automatically generating curricula for the target task in the form of Directed Acyclic Graphs (DAGs). AGCL encodes the specification in the form of a deterministic finite automaton (DFA), and then uses the DFA along with the Object-Oriented MDP (OOMDP) representation to generate a curriculum as a DAG, where the vertices correspond to tasks, and edges correspond to the direction of knowledge transfer. Experiments in gridworld and physics-based simulated robotics domains show that the curricula produced by AGCL achieve improved time-to-threshold performance on a complex sequential decision-making problem relative to state-of-the-art curriculum learning (e.g, teacher-student, self-play) and automaton-guided reinforcement learning baselines (e.g, Q-Learning for Reward Machines). Further, we demonstrate that AGCL performs well even in the presence of noise in the task's OOMDP description, and also when distractor objects are present that are not modeled in the logical specification of the tasks' objectives.

THIN: THrowable Information Networks and Application for Facial Expression Recognition in the Wild

For a number of machine learning problems, an exogenous variable can be identified such that it heavily influences the appearance of the different classes, and an ideal classifier should be invariant to this variable. An example of such exogenous variable is identity if facial expression recognition (FER) is considered. In this paper, we propose a dual exogenous/endogenous representation. The former captures the exogenous variable whereas the second one models the task at hand (e.g. facial expression). We design a prediction layer that uses a tree-gated deep ensemble conditioned by the exogenous representation. We also propose an exogenous dispelling loss to remove the exogenous information from the endogenous representation. Thus, the exogenous information is used two times in a throwable fashion, first as a conditioning variable for the target task, and second to create invariance within the endogenous representation. We call this method THIN, standing for THrowable Information Networks. We experimentally validate THIN in several contexts where an exogenous information can be identified, such as digit recognition under large rotations and shape recognition at multiple scales. We also apply it to FER with identity as the exogenous variable. We demonstrate that THIN significantly outperforms state-of-the-art approaches on several challenging datasets.

Label-Efficient Self-Supervised Federated Learning for Tackling Data Heterogeneity in Medical Imaging.

The collection and curation of large-scale medical datasets from multiple institutions is essential for training accurate deep learning models, but privacy concerns often hinder data sharing. Federated learning (FL) is a promising solution that enables privacy-preserving collaborative learning among different institutions, but it generally suffers from performance deterioration due to heterogeneous data distributions and a lack of quality labeled data. In this paper, we present a robust and label-efficient self-supervised FL framework for medical image analysis. Our method introduces a novel Transformer-based self-supervised pre-training paradigm that pre-trains models directly on decentralized target task datasets using masked image modeling, to facilitate more robust representation learning on heterogeneous data and effective knowledge transfer to downstream models. Extensive empirical results on simulated and real-world medical imaging non-IID federated datasets show that masked image modeling with Transformers significantly improves the robustness of models against various degrees of data heterogeneity. Notably, under severe data heterogeneity, our method, without relying on any additional pre-training data, achieves an improvement of 5.06%, 1.53% and 4.58% in test accuracy on retinal, dermatology and chest X-ray classification compared to the supervised baseline with ImageNet pre-training. In addition, we show that our federated self-supervised pre-training methods yield models that generalize better to out-of-distribution data and perform more effectively when fine-tuning with limited labeled data, compared to existing FL algorithms. The code is available at https://github.com/rui-yan/SSL-FL.