Due to increasing interest in adapting models on resource-constrained edges, parameter-efficient transfer learning has been widely explored. Among various methods, Visual Prompt Tuning (VPT), prepending learnable prompts to input space, shows competitive fine-tuning performance compared to training of full network parameters. However, VPT increases the number of input tokens, resulting in additional computational overhead. In this paper, we analyze the impact of the number of prompts on fine-tuning performance and self-attention operation in a vision transformer architecture. Through theoretical and empirical analysis we show that adding more prompts does not lead to linear performance improvement. Further, we propose a Prompt Condensation (PC) technique that aims to prevent performance degradation from using a small number of prompts. We validate our methods on FGVC and VTAB-1k tasks and show that our approach reduces the number of prompts by ~70% while maintaining accuracy.
Fine-tuning pre-trained language models for multiple tasks tends to be expensive in terms of storage. To mitigate this, parameter-efficient transfer learning (PETL) methods have been proposed to address this issue, but they still require a significant number of parameters and storage when being applied to broader ranges of tasks. To achieve even greater storage reduction, we propose PROPETL, a novel method that enables efficient sharing of a single PETL module which we call prototype network (e.g., adapter, LoRA, and prefix-tuning) across layers and tasks. We then learn binary masks to select different sub-networks from the shared prototype network and apply them as PETL modules into different layers. We find that the binary masks can determine crucial information from the network, which is often ignored in previous studies. Our work can also be seen as a type of pruning method, where we find that overparameterization also exists in the seemingly small PETL modules. We evaluate PROPETL on various downstream tasks and show that it can outperform other PETL methods with approximately 10% of the parameter storage required by the latter.
Deep learning (DL) has emerged as a powerful subset of machine learning (ML) and artificial intelligence (AI), outperforming traditional ML methods, especially in handling unstructured and large datasets. Its impact spans across various domains, including speech recognition, healthcare, autonomous vehicles, cybersecurity, predictive analytics, and more. However, the complexity and dynamic nature of real-world problems present challenges in designing effective deep learning models. Consequently, several deep learning models have been developed to address different problems and applications. In this article, we conduct a comprehensive survey of various deep learning models, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Generative Models, Deep Reinforcement Learning (DRL), and Deep Transfer Learning. We examine the structure, applications, benefits, and limitations of each model. Furthermore, we perform an analysis using three publicly available datasets: IMDB, ARAS, and Fruit-360. We compare the performance of six renowned deep learning models: CNN, Simple RNN, Long Short-Term Memory (LSTM), Bidirectional LSTM, Gated Recurrent Unit (GRU), and Bidirectional GRU.
Computational Pathology (CPATH) systems have the potential to automate diagnostic tasks. However, the artifacts on the digitized histological glass slides, known as Whole Slide Images (WSIs), may hamper the overall performance of CPATH systems. Deep Learning (DL) models such as Vision Transformers (ViTs) may detect and exclude artifacts before running the diagnostic algorithm. A simple way to develop robust and generalized ViTs is to train them on massive datasets. Unfortunately, acquiring large medical datasets is expensive and inconvenient, prompting the need for a generalized artifact detection method for WSIs. In this paper, we present a student-teacher recipe to improve the classification performance of ViT for the air bubbles detection task. ViT, trained under the student-teacher framework, boosts its performance by distilling existing knowledge from the high-capacity teacher model. Our best-performing ViT yields 0.961 and 0.911 F1-score and MCC, respectively, observing a 7% gain in MCC against stand-alone training. The proposed method presents a new perspective of leveraging knowledge distillation over transfer learning to encourage the use of customized transformers for efficient preprocessing pipelines in the CPATH systems.
This work presents a multi-resolution physics-informed recurrent neural network (MR PI-RNN), for simultaneous prediction of musculoskeletal (MSK) motion and parameter identification of the MSK systems. The MSK application was selected as the model problem due to its challenging nature in mapping the high-frequency surface electromyography (sEMG) signals to the low-frequency body joint motion controlled by the MSK and muscle contraction dynamics. The proposed method utilizes the fast wavelet transform to decompose the mixed frequency input sEMG and output joint motion signals into nested multi-resolution signals. The prediction model is subsequently trained on coarser-scale input-output signals using a gated recurrent unit (GRU), and then the trained parameters are transferred to the next level of training with finer-scale signals. These training processes are repeated recursively under a transfer-learning fashion until the full-scale training (i.e., with unfiltered signals) is achieved, while satisfying the underlying dynamic equilibrium. Numerical examples on recorded subject data demonstrate the effectiveness of the proposed framework in generating a physics-informed forward-dynamics surrogate, which yields higher accuracy in motion predictions of elbow flexion-extension of an MSK system compared to the case with single-scale training. The framework is also capable of identifying muscle parameters that are physiologically consistent with the subject's kinematics data.
In this paper, we consider the problem of learning a linear regression model on a data domain of interest (target) given few samples. To aid learning, we are provided with a set of pre-trained regression models that are trained on potentially different data domains (sources). Assuming a representation structure for the data generating linear models at the sources and the target domains, we propose a representation transfer based learning method for constructing the target model. The proposed scheme is comprised of two phases: (i) utilizing the different source representations to construct a representation that is adapted to the target data, and (ii) using the obtained model as an initialization to a fine-tuning procedure that re-trains the entire (over-parameterized) regression model on the target data. For each phase of the training method, we provide excess risk bounds for the learned model compared to the true data generating target model. The derived bounds show a gain in sample complexity for our proposed method compared to the baseline method of not leveraging source representations when achieving the same excess risk, therefore, theoretically demonstrating the effectiveness of transfer learning for linear regression.
Traditional biomedical artificial intelligence (AI) models, designed for specific tasks or modalities, often exhibit limited flexibility in real-world deployment and struggle to utilize holistic information. Generalist AI holds the potential to address these limitations due to its versatility in interpreting different data types and generating tailored outputs for diverse needs. However, existing biomedical generalist AI solutions are typically heavyweight and closed source to researchers, practitioners, and patients. Here, we propose BiomedGPT, the first open-source and lightweight vision-language foundation model, designed as a generalist capable of performing various biomedical tasks. BiomedGPT achieved state-of-the-art results in 16 out of 25 experiments while maintaining a computing-friendly model scale. We also conducted human evaluations to assess the capabilities of BiomedGPT in radiology visual question answering, report generation, and summarization. BiomedGPT exhibits robust prediction ability with a low error rate of 3.8% in question answering, satisfactory performance with an error rate of 8.3% in writing complex radiology reports, and competitive summarization ability with a nearly equivalent preference score to human experts. Our method demonstrates that effective training with diverse data can lead to more practical biomedical AI for improving diagnosis and workflow efficiency.
Holistic perception of affective attributes is an important human perceptual ability. However, this ability is far from being realized in current affective computing, as not all of the attributes are well studied and their interrelationships are poorly understood. In this work, we investigate the relationship between two affective attributes: personality and emotion, from a transfer learning perspective. Specifically, we transfer Transformer-based and wav2vec2-based emotion recognition models to perceive personality from speech across corpora. Compared with previous studies, our results show that transferring emotion recognition is effective for personality perception. Moreoever, this allows for better use and exploration of small personality corpora. We also provide novel findings on the relationship between personality and emotion that will aid future research on holistic affect recognition.
Knowledge graph completion (KGC), the task of predicting missing information based on the existing relational data inside a knowledge graph (KG), has drawn significant attention in recent years. However, the predictive power of KGC methods is often limited by the completeness of the existing knowledge graphs from different sources and languages. In monolingual and multilingual settings, KGs are potentially complementary to each other. In this paper, we study the problem of multi-KG completion, where we focus on maximizing the collective knowledge from different KGs to alleviate the incompleteness of individual KGs. Specifically, we propose a novel method called CKGC-CKD that uses relation-aware graph convolutional network encoder models on both individual KGs and a large fused KG in which seed alignments between KGs are regarded as edges for message propagation. An additional mutual knowledge distillation mechanism is also employed to maximize the knowledge transfer between the models of "global" fused KG and the "local" individual KGs. Experimental results on multilingual datasets have shown that our method outperforms all state-of-the-art models in the KGC task.
Transfer learning involves adapting a pre-trained model to novel downstream tasks. However, we observe that current transfer learning methods often fail to focus on task-relevant features. In this work, we explore refocusing model attention for transfer learning. We introduce Top-Down Attention Steering (TOAST), a novel transfer learning algorithm that keeps the pre-trained backbone frozen, selects task-relevant features in the output, and feeds those features back to the model to steer the attention to the task-specific features. By refocusing the attention only, TOAST achieves state-of-the-art results on a number of transfer learning benchmarks, while having a small number of tunable parameters. Compared to fully fine-tuning, LoRA, and prompt tuning, TOAST substantially improves performance across a range of fine-grained visual classification datasets (e.g., 81.1% -> 86.2% on FGVC). TOAST also outperforms the fully fine-tuned Alpaca and Vicuna models on instruction-following language generation. Code is available at https://github.com/bfshi/TOAST.
Designing novel functional proteins remains a slow and expensive process due to a variety of protein engineering challenges; in particular, the number of protein variants that can be experimentally tested in a given assay pales in comparison to the vastness of the overall sequence space, resulting in low hit rates and expensive wet lab testing cycles. In this paper, we propose a few-shot learning approach to novel protein design that aims to accelerate the expensive wet lab testing cycle and is capable of leveraging a training dataset that is both small and skewed ($\approx 10^5$ datapoints, $< 1\%$ positive hits). Our approach is composed of two parts: a semi-supervised transfer learning approach to generate a discrete fitness landscape for a desired protein function and a novel evolutionary Monte Carlo Markov Chain sampling algorithm to more efficiently explore the fitness landscape. We demonstrate the performance of our approach by experimentally screening predicted high fitness gene activators, resulting in a dramatically improved hit rate compared to existing methods. Our method can be easily adapted to other protein engineering and design problems, particularly where the cost associated with obtaining labeled data is significantly high. We have provided open source code for our method at https:// github.com/SuperSecretBioTech/evolutionary_monte_carlo_search.
Real-world multinational e-commerce companies, such as Amazon and eBay, serve in multiple countries and regions. Some markets are data-scarce, while others are data-rich. In recent years, cross-market recommendation (XMR) has been proposed to bolster data-scarce markets by leveraging auxiliary information from data-rich markets. Previous XMR algorithms have employed techniques such as sharing bottom or incorporating inter-market similarity to optimize the performance of XMR. However, the existing approaches suffer from two crucial limitations: (1) They ignore the co-occurrences of items provided by data-rich markets. (2) They do not adequately tackle the issue of negative transfer stemming from disparities across diverse markets. In order to address these limitations, we propose a novel session-based model called Bert4XMR, which is able to model item co-occurrences across markets and mitigate negative transfer. Specifically, we employ the pre-training and fine-tuning paradigm to facilitate knowledge transfer across markets. Pre-training occurs on global markets to learn item co-occurrences, while fine-tuning happens in the target market for model customization. To mitigate potential negative transfer, we separate the item representations into market embeddings and item embeddings. Market embeddings model the bias associated with different markets, while item embeddings learn generic item representations. Extensive experiments conducted on seven real-world datasets illustrate our model's effectiveness. It outperforms the suboptimal model by an average of $4.82\%$, $4.73\%$, $7.66\%$, and $6.49\%$ across four metrics. Through the ablation study, we experimentally demonstrate that the market embedding approach helps prevent negative transfer, especially in data-scarce markets. Our implementations are available at https://github.com/laowangzi/Bert4XMR.
Adapters, a plug-in neural network module with some tunable parameters, have emerged as a parameter-efficient transfer learning technique for adapting pre-trained models to downstream tasks, especially for natural language processing (NLP) and computer vision (CV) fields. Meanwhile, learning recommendation models directly from raw item modality features -- e.g., texts of NLP and images of CV -- can enable effective and transferable recommender systems (called TransRec). In view of this, a natural question arises: can adapter-based learning techniques achieve parameter-efficient TransRec with good performance? To this end, we perform empirical studies to address several key sub-questions. First, we ask whether the adapter-based TransRec performs comparably to TransRec based on standard full-parameter fine-tuning? does it hold for recommendation with different item modalities, e.g., textual RS and visual RS. If yes, we benchmark these existing adapters, which have been shown to be effective in NLP and CV tasks, in item recommendation tasks. Third, we carefully study several key factors for the adapter-based TransRec in terms of where and how to insert these adapters? Finally, we look at the effects of adapter-based TransRec by either scaling up its source training data or scaling down its target training data. Our paper provides key insights and practical guidance on unified & transferable recommendation -- a less studied recommendation scenario. We release our codes and other materials at: https://github.com/westlake-repl/Adapter4Rec/.
Question-answering (QA) tasks often investigate specific question types, knowledge domains, or reasoning skills, leading to specialized models catering to specific categories of QA tasks. While recent research has explored the idea of unified QA models, such models are usually explored for high-resource scenarios and require re-training to extend their capabilities. To overcome these drawbacks, the paper explores the potential of two paradigms of tuning, model, and prompts, for unified QA under a low-resource setting. The paper provides an exhaustive analysis of their applicability using 16 QA datasets, revealing that prompt tuning can perform as well as model tuning in a few-shot setting with a good initialization. The study also shows that parameter-sharing results in superior few-shot performance, simple knowledge transfer techniques for prompt initialization can be effective, and prompt tuning achieves a significant performance boost from pre-training in a low-resource regime. The research offers insights into the advantages and limitations of prompt tuning for unified QA in a few-shot setting, contributing to the development of effective and efficient systems in low-resource scenarios.
Image segmentation is a branch of computer vision that is widely used in real world applications including biomedical image processing. With recent advancement of deep learning, image segmentation has achieved at a very high level performance. Recently, UNet architecture is found as the core of novel deep learning segmentation methods. In this paper we implement UNet architecture from scratch with using basic blocks in Pytorch and evaluate its performance on multiple biomedical image datasets. We also use transfer learning to apply novel modified UNet segmentation packages on the biomedical image datasets. We fine tune the pre-trained transferred model with each specific dataset. We compare its performance with our fundamental UNet implementation. We show that transferred learning model has better performance in image segmentation than UNet model that is implemented from scratch.
Applications of deep learning in financial market prediction has attracted huge attention from investors and researchers. In particular, intra-day prediction at the minute scale, the dramatically fluctuating volume and stock prices within short time periods have posed a great challenge for the convergence of networks result. Informer is a more novel network, improved on Transformer with smaller computational complexity, longer prediction length and global time stamp features. We have designed three experiments to compare Informer with the commonly used networks LSTM, Transformer and BERT on 1-minute and 5-minute frequencies for four different stocks/ market indices. The prediction results are measured by three evaluation criteria: MAE, RMSE and MAPE. Informer has obtained best performance among all the networks on every dataset. Network without the global time stamp mechanism has significantly lower prediction effect compared to the complete Informer; it is evident that this mechanism grants the time series to the characteristics and substantially improves the prediction accuracy of the networks. Finally, transfer learning capability experiment is conducted, Informer also achieves a good performance. Informer has good robustness and improved performance in market prediction, which can be exactly adapted to real trading.
Fine-tuning large-scale Transformers has led to the explosion of many AI applications across Natural Language Processing and Computer Vision tasks. However, fine-tuning all pre-trained model parameters becomes impractical as the model size and number of tasks increase. Parameter-efficient transfer learning (PETL) methods aim to address these challenges. While effective in reducing the number of trainable parameters, PETL methods still require significant energy and computational resources to fine-tune. In this paper, we introduce \textbf{RE}current \textbf{AD}aption (READ) -- a lightweight and memory-efficient fine-tuning method -- to overcome the limitations of the current PETL approaches. Specifically, READ inserts a small RNN network alongside the backbone model so that the model does not have to back-propagate through the large backbone network. Through comprehensive empirical evaluation of the GLUE benchmark, we demonstrate READ can achieve a $56\%$ reduction in the training memory consumption and an $84\%$ reduction in the GPU energy usage while retraining high model quality compared to full-tuning. Additionally, the model size of READ does not grow with the backbone model size, making it a highly scalable solution for fine-tuning large Transformers.
Training offline RL models using visual inputs poses two significant challenges, i.e., the overfitting problem in representation learning and the overestimation bias for expected future rewards. Recent work has attempted to alleviate the overestimation bias by encouraging conservative behaviors. This paper, in contrast, tries to build more flexible constraints for value estimation without impeding the exploration of potential advantages. The key idea is to leverage off-the-shelf RL simulators, which can be easily interacted with in an online manner, as the "test bed" for offline policies. To enable effective online-to-offline knowledge transfer, we introduce CoWorld, a model-based RL approach that mitigates cross-domain discrepancies in state and reward spaces. Experimental results demonstrate the effectiveness of CoWorld, outperforming existing RL approaches by large margins.
The human brain is the central hub of the neurobiological system, controlling behavior and cognition in complex ways. Recent advances in neuroscience and neuroimaging analysis have shown a growing interest in the interactions between brain regions of interest (ROIs) and their impact on neural development and disorder diagnosis. As a powerful deep model for analyzing graph-structured data, Graph Neural Networks (GNNs) have been applied for brain network analysis. However, training deep models requires large amounts of labeled data, which is often scarce in brain network datasets due to the complexities of data acquisition and sharing restrictions. To make the most out of available training data, we propose PTGB, a GNN pre-training framework that captures intrinsic brain network structures, regardless of clinical outcomes, and is easily adaptable to various downstream tasks. PTGB comprises two key components: (1) an unsupervised pre-training technique designed specifically for brain networks, which enables learning from large-scale datasets without task-specific labels; (2) a data-driven parcellation atlas mapping pipeline that facilitates knowledge transfer across datasets with different ROI systems. Extensive evaluations using various GNN models have demonstrated the robust and superior performance of PTGB compared to baseline methods.
Voice technology has become ubiquitous recently. However, the accuracy, and hence experience, in different languages varies significantly, which makes the technology not equally inclusive. The availability of data for different languages is one of the key factors affecting accuracy, especially in training of all-neural end-to-end automatic speech recognition systems. Cross-lingual knowledge transfer and iterative pseudo-labeling are two techniques that have been shown to be successful for improving the accuracy of ASR systems, in particular for low-resource languages, like Ukrainian. Our goal is to train an all-neural Transducer-based ASR system to replace a DNN-HMM hybrid system with no manually annotated training data. We show that the Transducer system trained using transcripts produced by the hybrid system achieves 18% reduction in terms of word error rate. However, using a combination of cross-lingual knowledge transfer from related languages and iterative pseudo-labeling, we are able to achieve 35% reduction of the error rate.