Labeled data are critical to modern machine learning applications, but obtaining labels can be expensive. To mitigate this cost, machine learning methods, such as transfer learning, semi-supervised learning and active learning, aim to be label-efficient: achieving high predictive performance from relatively few labeled examples. While obtaining the best label-efficiency in practice often requires combinations of these techniques, existing benchmark and evaluation frameworks do not capture a concerted combination of all such techniques. This paper addresses this deficiency by introducing LabelBench, a new computationally-efficient framework for joint evaluation of multiple label-efficient learning techniques. As an application of LabelBench, we introduce a novel benchmark of state-of-the-art active learning methods in combination with semi-supervised learning for fine-tuning pretrained vision transformers. Our benchmark demonstrates better label-efficiencies than previously reported in active learning. LabelBench's modular codebase is open-sourced for the broader community to contribute label-efficient learning methods and benchmarks. The repository can be found at: https://github.com/EfficientTraining/LabelBench.
Pre-training & fine-tuning is a prevalent paradigm in computer vision (CV). Recently, parameter-efficient transfer learning (PETL) methods have shown promising performance in adapting to downstream tasks with only a few trainable parameters. Despite their success, the existing PETL methods in CV can be computationally expensive and require large amounts of memory and time cost during training, which limits low-resource users from conducting research and applications on large models. In this work, we propose Parameter, Memory, and Time Efficient Visual Adapter ($\mathrm{E^3VA}$) tuning to address this issue. We provide a gradient backpropagation highway for low-rank adapters which eliminates the need for expensive backpropagation through the frozen pre-trained model, resulting in substantial savings of training memory and training time. Furthermore, we optimise the $\mathrm{E^3VA}$ structure for CV tasks to promote model performance. Extensive experiments on COCO, ADE20K, and Pascal VOC benchmarks show that $\mathrm{E^3VA}$ can save up to 62.2% training memory and 26.2% training time on average, while achieving comparable performance to full fine-tuning and better performance than most PETL methods. Note that we can even train the Swin-Large-based Cascade Mask RCNN on GTX 1080Ti GPUs with less than 1.5% trainable parameters.
Document understanding tasks, in particular, Visually-rich Document Entity Retrieval (VDER), have gained significant attention in recent years thanks to their broad applications in enterprise AI. However, publicly available data have been scarce for these tasks due to strict privacy constraints and high annotation costs. To make things worse, the non-overlapping entity spaces from different datasets hinder the knowledge transfer between document types. In this paper, we propose a method to collect massive-scale and weakly labeled data from the web to benefit the training of VDER models. The collected dataset, named DocumentNet, does not depend on specific document types or entity sets, making it universally applicable to all VDER tasks. The current DocumentNet consists of 30M documents spanning nearly 400 document types organized in a four-level ontology. Experiments on a set of broadly adopted VDER tasks show significant improvements when DocumentNet is incorporated into the pre-training for both classic and few-shot learning settings. With the recent emergence of large language models (LLMs), DocumentNet provides a large data source to extend their multi-modal capabilities for VDER.
Over the past decade, deep learning research has been accelerated by increasingly powerful hardware, which facilitated rapid growth in the model complexity and the amount of data ingested. This is becoming unsustainable and therefore refocusing on efficiency is necessary. In this paper, we employ transfer learning to improve training efficiency for large-scale spatial problems. We propose that a convolutional neural network (CNN) can be trained on small windows of signals, but evaluated on arbitrarily large signals with little to no performance degradation, and provide a theoretical bound on the resulting generalization error. Our proof leverages shift-equivariance of CNNs, a property that is underexploited in transfer learning. The theoretical results are experimentally supported in the context of mobile infrastructure on demand (MID). The proposed approach is able to tackle MID at large scales with hundreds of agents, which was computationally intractable prior to this work.
There are numerous advantages of deep neural network surrogate modeling for response time-history prediction. However, due to the high cost of refined numerical simulations and actual experiments, the lack of data has become an unavoidable bottleneck in practical applications. An iterative self-transfer learningmethod for training neural networks based on small datasets is proposed in this study. A new mapping-based transfer learning network, named as deep adaptation network with three branches for regression (DAN-TR), is proposed. A general iterative network training strategy is developed by coupling DAN-TR and the pseudo-label strategy, and the establishment of corresponding datasets is also discussed. Finally, a complex component is selected as a case study. The results show that the proposed method can improve the model performance by near an order of magnitude on small datasets without the need of external labeled samples,well behaved pre-trainedmodels, additional artificial labeling, and complex physical/mathematical analysis.
We propose a novel framework and a solution to tackle the continual learning (CL) problem with changing network architectures. Most CL methods focus on adapting a single architecture to a new task/class by modifying its weights. However, with rapid progress in architecture design, the problem of adapting existing solutions to novel architectures becomes relevant. To address this limitation, we propose Heterogeneous Continual Learning (HCL), where a wide range of evolving network architectures emerge continually together with novel data/tasks. As a solution, we build on top of the distillation family of techniques and modify it to a new setting where a weaker model takes the role of a teacher; meanwhile, a new stronger architecture acts as a student. Furthermore, we consider a setup of limited access to previous data and propose Quick Deep Inversion (QDI) to recover prior task visual features to support knowledge transfer. QDI significantly reduces computational costs compared to previous solutions and improves overall performance. In summary, we propose a new setup for CL with a modified knowledge distillation paradigm and design a quick data inversion method to enhance distillation. Our evaluation of various benchmarks shows a significant improvement on accuracy in comparison to state-of-the-art methods over various networks architectures.
Deep learning is the state-of-the-art for medical imaging tasks, but requires large, labeled datasets. For risk prediction, large datasets are rare since they require both imaging and follow-up (e.g., diagnosis codes). However, the release of publicly available imaging data with diagnostic labels presents an opportunity for self and semi-supervised approaches to improve label efficiency for risk prediction. Though several studies have compared self-supervised approaches in natural image classification, object detection, and medical image interpretation, there is limited data on which approaches learn robust representations for risk prediction. We present a comparison of semi- and self-supervised learning to predict mortality risk using chest x-ray images. We find that a semi-supervised autoencoder outperforms contrastive and transfer learning in internal and external validation.
Personalized dialogue agents (DAs) powered by large pre-trained language models (PLMs) often rely on explicit persona descriptions to maintain personality consistency. However, such descriptions may not always be available or may pose privacy concerns. To tackle this bottleneck, we introduce PersonaPKT, a lightweight transfer learning approach that can build persona-consistent dialogue models without explicit persona descriptions. By representing each persona as a continuous vector, PersonaPKT learns implicit persona-specific features directly from a small number of dialogue samples produced by the same persona, adding less than 0.1% trainable parameters for each persona on top of the PLM backbone. Empirical results demonstrate that PersonaPKT effectively builds personalized DAs with high storage efficiency, outperforming various baselines in terms of persona consistency while maintaining good response generation quality. In addition, it enhances privacy protection by avoiding explicit persona descriptions. Overall, PersonaPKT is an effective solution for creating personalized DAs that respect user privacy.
Medical image segmentation based on deep learning often fails when deployed on images from a different domain. The domain adaptation methods aim to solve domain-shift challenges, but still face some problems. The transfer learning methods require annotation on the target domain, and the generative unsupervised domain adaptation (UDA) models ignore domain-specific representations, whose generated quality highly restricts segmentation performance. In this study, we propose a novel Structure-Modal Constrained (SMC) UDA framework based on a discriminative paradigm and introduce edge structure as a bridge between domains. The proposed multi-modal learning backbone distills structure information from image texture to distinguish domain-invariant edge structure. With the structure-constrained self-learning and progressive ROI, our methods segment the kidney by locating the 3D spatial structure of the edge. We evaluated SMC-UDA on public renal segmentation datasets, adapting from the labeled source domain (CT) to the unlabeled target domain (CT/MRI). The experiments show that our proposed SMC-UDA has a strong generalization and outperforms generative UDA methods.
The utilization of clinical reports for various secondary purposes, including health research and treatment monitoring, is crucial for enhancing patient care. Natural Language Processing (NLP) tools have emerged as valuable assets for extracting and processing relevant information from these reports. However, the availability of specialized language models for the clinical domain in Spanish has been limited. In this paper, we introduce EriBERTa, a bilingual domain-specific language model pre-trained on extensive medical and clinical corpora. We demonstrate that EriBERTa outperforms previous Spanish language models in the clinical domain, showcasing its superior capabilities in understanding medical texts and extracting meaningful information. Moreover, EriBERTa exhibits promising transfer learning abilities, allowing for knowledge transfer from one language to another. This aspect is particularly beneficial given the scarcity of Spanish clinical data.
We present Generalized LoRA (GLoRA), an advanced approach for universal parameter-efficient fine-tuning tasks. Enhancing Low-Rank Adaptation (LoRA), GLoRA employs a generalized prompt module to optimize pre-trained model weights and adjust intermediate activations, providing more flexibility and capability across diverse tasks and datasets. Moreover, GLoRA facilitates efficient parameter adaptation by employing a scalable, modular, layer-wise structure search that learns individual adapter of each layer. Originating from a unified mathematical formulation, GLoRA exhibits strong transfer learning, few-shot learning and domain generalization abilities, as it adapts to new tasks through not only weights but also additional dimensions like activations. Comprehensive experiments demonstrate that GLoRA outperforms all previous methods in natural, specialized, and structured vision benchmarks, achieving superior accuracy with fewer parameters and computations. The proposed method on LLaMA-1 and LLaMA-2 also show considerable enhancements compared to the original LoRA in the language domain. Furthermore, our structural re-parameterization design ensures that GLoRA incurs no extra inference cost, rendering it a practical solution for resource-limited applications. Code and models are available at: https://github.com/Arnav0400/ViT-Slim/tree/master/GLoRA.
This article investigates the knowledge transfer from the RuQTopics dataset. This Russian topical dataset combines a large sample number (361,560 single-label, 170,930 multi-label) with extensive class coverage (76 classes). We have prepared this dataset from the "Yandex Que" raw data. By evaluating the RuQTopics - trained models on the six matching classes of the Russian MASSIVE subset, we have proved that the RuQTopics dataset is suitable for real-world conversational tasks, as the Russian-only models trained on this dataset consistently yield an accuracy around 85\% on this subset. We also have figured out that for the multilingual BERT, trained on the RuQTopics and evaluated on the same six classes of MASSIVE (for all MASSIVE languages), the language-wise accuracy closely correlates (Spearman correlation 0.773 with p-value 2.997e-11) with the approximate size of the pretraining BERT's data for the corresponding language. At the same time, the correlation of the language-wise accuracy with the linguistical distance from Russian is not statistically significant.
Deep learning (DL) models have seen increased attention for time series forecasting, yet the application on cyber-physical systems (CPS) is hindered by the lacking robustness of these methods. Thus, this study evaluates the robustness and generalization performance of DL architectures on multivariate time series data from CPS. Our investigation focuses on the models' ability to handle a range of perturbations, such as sensor faults and noise, and assesses their impact on overall performance. Furthermore, we test the generalization and transfer learning capabilities of these models by exposing them to out-of-distribution (OOD) samples. These include deviations from standard system operations, while the core dynamics of the underlying physical system are preserved. Additionally, we test how well the models respond to several data augmentation techniques, including added noise and time warping. Our experimental framework utilizes a simulated three-tank system, proposed as a novel benchmark for evaluating the robustness and generalization performance of DL algorithms in CPS data contexts. The findings reveal that certain DL model architectures and training techniques exhibit superior effectiveness in handling OOD samples and various perturbations. These insights have significant implications for the development of DL models that deliver reliable and robust performance in real-world CPS applications.
Open data is frequently released spatially aggregated, usually to comply with privacy policies. But coarse, heterogeneous aggregations complicate learning and integration for downstream AI/ML systems. In this work, we consider models to disaggregate spatio-temporal data from a low-resolution, irregular partition (e.g., census tract) to a high-resolution, irregular partition (e.g., city block). We propose an overarching model named the Structurally-Aware Recurrent Network (SARN), which integrates structurally-aware spatial attention (SASA) layers into the Gated Recurrent Unit (GRU) model. The spatial attention layers capture spatial interactions among regions, while the gated recurrent module captures the temporal dependencies. Each SASA layer calculates both global and structural attention -- global attention facilitates comprehensive interactions between different geographic levels, while structural attention leverages the containment relationship between different geographic levels (e.g., a city block being wholly contained within a census tract) to ensure coherent and consistent results. For scenarios with limited historical training data, we explore transfer learning and show that a model pre-trained on one city variable can be fine-tuned for another city variable using only a few hundred samples. Evaluating these techniques on two mobility datasets, we find that on both datasets, SARN significantly outperforms other neural models (5% and 1%) and typical heuristic methods (40% and 14%), enabling us to generate realistic, high-quality fine-grained data for downstream applications.
Potatoes are the third-largest food crop globally, but their production frequently encounters difficulties because of aggressive pest infestations. The aim of this study is to investigate the various types and characteristics of these pests and propose an efficient PotatoPestNet AI-based automatic potato pest identification system. To accomplish this, we curated a reliable dataset consisting of eight types of potato pests. We leveraged the power of transfer learning by employing five customized, pre-trained transfer learning models: CMobileNetV2, CNASLargeNet, CXception, CDenseNet201, and CInceptionV3, in proposing a robust PotatoPestNet model to accurately classify potato pests. To improve the models' performance, we applied various augmentation techniques, incorporated a global average pooling layer, and implemented proper regularization methods. To further enhance the performance of the models, we utilized random search (RS) optimization for hyperparameter tuning. This optimization technique played a significant role in fine-tuning the models and achieving improved performance. We evaluated the models both visually and quantitatively, utilizing different evaluation metrics. The robustness of the models in handling imbalanced datasets was assessed using the Receiver Operating Characteristic (ROC) curve. Among the models, the Customized Tuned Inception V3 (CTInceptionV3) model, optimized through random search, demonstrated outstanding performance. It achieved the highest accuracy (91%), precision (91%), recall (91%), and F1-score (91%), showcasing its superior ability to accurately identify and classify potato pests.
In dysarthric speech recognition, data scarcity and the vast diversity between dysarthric speakers pose significant challenges. While finetuning has been a popular solution, it can lead to overfitting and low parameter efficiency. Adapter modules offer a better solution, with their small size and easy applicability. Additionally, Adapter Fusion can facilitate knowledge transfer from multiple learned adapters, but may employ more parameters. In this work, we apply Adapter Fusion for target speaker adaptation and speech recognition, achieving acceptable accuracy with significantly fewer speaker-specific trainable parameters than classical finetuning methods. We further improve the parameter efficiency of the fusion layer by reducing the size of query and key layers and using Householder transformation to reparameterize the value linear layer. Our proposed fusion layer achieves comparable recognition results to the original method with only one third of the parameters.
Hypernetworks, or hypernets for short, are neural networks that generate weights for another neural network, known as the target network. They have emerged as a powerful deep learning technique that allows for greater flexibility, adaptability, dynamism, faster training, information sharing, and model compression. Hypernets have shown promising results in a variety of deep learning problems, including continual learning, causal inference, transfer learning, weight pruning, uncertainty quantification, zero-shot learning, natural language processing, and reinforcement learning. Despite their success across different problem settings, there is currently no comprehensive review available to inform researchers about the latest developments and to assist in utilizing hypernets. To fill this gap, we review the progress in hypernets. We present an illustrative example of training deep neural networks using hypernets and propose categorizing hypernets based on five design criteria: inputs, outputs, variability of inputs and outputs, and the architecture of hypernets. We also review applications of hypernets across different deep learning problem settings, followed by a discussion of general scenarios where hypernets can be effectively employed. Finally, we discuss the challenges and future directions that remain underexplored in the field of hypernets. We believe that hypernetworks have the potential to revolutionize the field of deep learning. They offer a new way to design and train neural networks, and they have the potential to improve the performance of deep learning models on a variety of tasks. Through this review, we aim to inspire further advancements in deep learning through hypernetworks.
Transfer learning is a valuable tool in deep learning as it allows propagating information from one "source dataset" to another "target dataset", especially in the case of a small number of training examples in the latter. Yet, discrepancies between the underlying distributions of the source and target data are commonplace and are known to have a substantial impact on algorithm performance. In this work we suggest novel information-theoretic approaches for the analysis of the performance of deep neural networks in the context of transfer learning. We focus on the task of semi-supervised transfer learning, in which unlabeled samples from the target dataset are available during network training on the source dataset. Our theory suggests that one may improve the transferability of a deep neural network by incorporating regularization terms on the target data based on information-theoretic quantities, namely the Mutual Information and the Lautum Information. We demonstrate the effectiveness of the proposed approaches in various semi-supervised transfer learning experiments.
Named Entity Recognition (NER) is a fundamental task in NLP that is used to locate the key information in text and is primarily applied in conversational and search systems. In commercial applications, NER or comparable slot-filling methods have been widely deployed for popular languages. NER is used in applications such as human resources, customer service, search engines, content classification, and academia. In this paper, we draw focus on identifying name entities for low-resource Indian languages that are closely related, like Hindi and Marathi. We use various adaptations of BERT such as baseBERT, AlBERT, and RoBERTa to train a supervised NER model. We also compare multilingual models with monolingual models and establish a baseline. In this work, we show the assisting capabilities of the Hindi and Marathi languages for the NER task. We show that models trained using multiple languages perform better than a single language. However, we also observe that blind mixing of all datasets doesn't necessarily provide improvements and data selection methods may be required.
Image Augmentations are widely used to reduce overfitting in neural networks. However, the explainability of their benefits largely remains a mystery. We study which layers of residual neural networks (ResNets) are most affected by augmentations using Centered Kernel Alignment (CKA). We do so by analyzing models of varying widths and depths, as well as whether their weights are initialized randomly or through transfer learning. We find that the pattern of how the layers are affected depends on the model's depth, and that networks trained with augmentation that use information from two images affect the learned weights significantly more than augmentations that operate on a single image. Deeper layers of ResNets initialized with ImageNet-1K weights and fine-tuned receive more impact from the augmentations than early layers. Understanding the effects of image augmentations on CNNs will have a variety of applications, such as determining how far back one needs to fine-tune a network and which layers should be frozen when implementing layer freezing algorithms.