In natural language processing (NLP), code-mixing (CM) is a challenging task, especially when the mixed languages include dialects. In Southeast Asian countries such as Singapore, Indonesia, and Malaysia, Hokkien-Mandarin is the most widespread code-mixed language pair among Chinese immigrants, and it is also common in Taiwan. However, dialects such as Hokkien often have a scarcity of resources and the lack of an official writing system, limiting the development of dialect CM research. In this paper, we propose a method to construct a Hokkien-Mandarin CM dataset to mitigate the limitation, overcome the morphological issue under the Sino-Tibetan language family, and offer an efficient Hokkien word segmentation method through a linguistics-based toolkit. Furthermore, we use our proposed dataset and employ transfer learning to train the XLM (cross-lingual language model) for translation tasks. To fit the code-mixing scenario, we adapt XLM slightly. We found that by using linguistic knowledge, rules, and language tags, the model produces good results on CM data translation while maintaining monolingual translation quality.
Most Transformer language models are primarily pretrained on English text, limiting their use for other languages. As the model sizes grow, the performance gap between English and other languages with fewer compute and data resources increases even further. Consequently, more resource-efficient training methods are needed to bridge the gap for languages with fewer resources available. To address this problem, we introduce a cross-lingual and progressive transfer learning approach, called CLP-Transfer, that transfers models from a source language, for which pretrained models are publicly available, like English, to a new target language. As opposed to prior work, which focused on the cross-lingual transfer between two languages, we extend the transfer to the model size. Given a pretrained model in a source language, we aim for a same-sized model in a target language. Instead of training a model from scratch, we exploit a smaller model that is in the target language but requires much fewer resources. Both small and source models are then used to initialize the token embeddings of the larger model based on the overlapping vocabulary of the source and target language. All remaining weights are reused from the model in the source language. This approach outperforms the sole cross-lingual transfer and can save up to 80% of the training steps compared to the random initialization.
By promising more accurate diagnostics and individual treatment recommendations, deep neural networks and in particular convolutional neural networks have advanced to a powerful tool in medical imaging. Here, we first give an introduction into methodological key concepts and resulting methodological promises including representation and transfer learning, as well as modelling domain-specific priors. After reviewing recent applications within neuroimaging-based psychiatric research, such as the diagnosis of psychiatric diseases, delineation of disease subtypes, normative modeling, and the development of neuroimaging biomarkers, we discuss current challenges. This includes for example the difficulty of training models on small, heterogeneous and biased data sets, the lack of validity of clinical labels, algorithmic bias, and the influence of confounding variables.
People all over the globe are affected by pneumonia but deaths due to it are highest in Sub-Saharan Asia and South Asia. In recent years, the overall incidence and mortality rate of pneumonia regardless of the utilization of effective vaccines and compelling antibiotics has escalated. Thus, pneumonia remains a disease that needs spry prevention and treatment. The widespread prevalence of pneumonia has caused the research community to come up with a framework that helps detect, diagnose and analyze diseases accurately and promptly. One of the major hurdles faced by the Artificial Intelligence (AI) research community is the lack of publicly available datasets for chest diseases, including pneumonia . Secondly, few of the available datasets are highly imbalanced (normal examples are over sampled, while samples with ailment are in severe minority) making the problem even more challenging. In this article we present a novel framework for the detection of pneumonia. The novelty of the proposed methodology lies in the tackling of class imbalance problem. The Generative Adversarial Network (GAN), specifically a combination of Deep Convolutional Generative Adversarial Network (DCGAN) and Wasserstein GAN gradient penalty (WGAN-GP) was applied on the minority class ``Pneumonia'' for augmentation, whereas Random Under-Sampling (RUS) was done on the majority class ``No Findings'' to deal with the imbalance problem. The ChestX-Ray8 dataset, one of the biggest datasets, is used to validate the performance of the proposed framework. The learning phase is completed using transfer learning on state-of-the-art deep learning models i.e. ResNet-50, Xception, and VGG-16. Results obtained exceed state-of-the-art.
In this paper, we present StEduCov, an annotated dataset for the analysis of stances toward online education during the COVID-19 pandemic. StEduCov consists of 16,572 tweets gathered over 15 months, from March 2020 to May 2021, using the Twitter API. The tweets were manually annotated into the classes agree, disagreeor neutral. We performed benchmarking on the dataset using state-of-the-art and traditional machine learning models. Specifically, we trained deep learning models—bidirectional encoder representations from transformers, long short-term memory, convolutional neural networks, attention-based biLSTM and Naive Bayes SVM—in addition to naive Bayes, logistic regression, support vector machines, decision trees, K-nearest neighbor and random forest. The average accuracy in the 10-fold cross-validation of these models ranged from 75% to 84.8% and from 52.6% to 68% for binary and multi-class stance classifications, respectively. Performances were affected by high vocabulary overlaps between classes and unreliable transfer learning using deep models pre-trained on general texts in relation to specific domains such as COVID-19 and distance education.
Due to the high temporal uncertainty and low signal-to-noise ratio, transfer learning for univariate time series forecasting remains a challenging task. In addition, data scarcity, which is commonly encountered in business forecasting, further limits the application of conventional transfer learning protocols. In this work, we have developed, DATSING, a transfer learning-based framework that effectively leverages cross-domain time series latent representations to augment target domain forecasting. In particular, we aim to transfer domain-invariant feature representations from a pre-trained stacked deep residual network to the target domains, so as to assist the prediction of each target time series. To effectively avoid noisy feature representations, we propose a two-phased framework which first clusters similar mixed domains time series data and then performs a fine-tuning procedure with domain adversarial regularization to achieve better out-of-sample generalization. Extensive experiments with real-world datasets have demonstrated that our method significantly improves the forecasting performance of the pre-trained model. DATSING has the unique potential to empower forecasting practitioners to unleash the power of cross-domain time series data.
State-of-the-art video-text retrieval (VTR) methods typically involve fully fine-tuning a pre-trained model (e.g. CLIP) on specific datasets. However, this can result in significant storage costs in practical applications as a separate model per task must be stored. To address this issue, we present our pioneering work that enables parameter-efficient VTR using a pre-trained model, with only a small number of tunable parameters during training. Towards this goal, we propose a new method dubbed Multimodal Video Adapter (MV-Adapter) for efficiently transferring the knowledge in the pre-trained CLIP from image-text to video-text. Specifically, MV-Adapter utilizes bottleneck structures in both video and text branches, along with two novel components. The first is a Temporal Adaptation Module that is incorporated in the video branch to introduce global and local temporal contexts. We also train weights calibrations to adjust to dynamic variations across frames. The second is Cross Modality Tying that generates weights for video/text branches through sharing cross modality factors, for better aligning between modalities. Thanks to above innovations, MV-Adapter can achieve comparable or better performance than standard full fine-tuning with negligible parameters overhead. Notably, MV-Adapter consistently outperforms various competing methods in V2T/T2V tasks with large margins on five widely used VTR benchmarks (MSR-VTT, MSVD, LSMDC, DiDemo, and ActivityNet).
In this paper, we propose to exploit the side-tuning framework for multimodal document classification. Side-tuning is a methodology for network adaptation recently introduced to solve some of the problems related to previous approaches. Thanks to this technique it is actually possible to overcome model rigidity and catastrophic forgetting of transfer learning by fine-tuning. The proposed solution uses off-the-shelf deep learning architectures leveraging the side-tuning framework to combine a base model with a tandem of two side networks. We show that side-tuning can be successfully employed also when different data sources are considered, e.g. text and images in document classification. The experimental results show that this approach pushes further the limit for document classification accuracy with respect to the state of the art.
Two-dimensional (2D) materials have wide applications in superconductors, quantum, and topological materials. However, their rational design is not well established, and currently less than 6,000 experimentally synthesized 2D materials have been reported. Recently, deep learning, data-mining, and density functional theory (DFT)-based high-throughput calculations are widely performed to discover potential new materials for diverse applications. Here we propose a generative material design pipeline, namely material transformer generator(MTG), for large-scale discovery of hypothetical 2D materials. We train two 2D materials composition generators using self-learning neural language models based on Transformers with and without transfer learning. The models are then used to generate a large number of candidate 2D compositions, which are fed to known 2D materials templates for crystal structure prediction. Next, we performed DFT computations to study their thermodynamic stability based on energy-above-hull and formation energy. We report four new DFT-verified stable 2D materials with zero e-above-hull energies, including NiCl$_4$, IrSBr, CuBr$_3$, and CoBrCl. Our work thus demonstrates the potential of our MTG generative materials design pipeline in the discovery of novel 2D materials and other functional materials.
There are many computer vision applications including object segmentation, classification, object detection, and reconstruction for which machine learning (ML) shows state-of-the-art performance. Nowadays, we can build ML tools for such applications with real-world accuracy. However, each tool works well within the domain in which it has been trained and developed. Often, when we train a model on a dataset in one specific domain and test on another unseen domain known as an out of distribution (OOD) dataset, models or ML tools show a decrease in performance. For instance, when we train a simple classifier on real-world images and apply that model on the same classes but with a different domain like cartoons, paintings or sketches then the performance of ML tools disappoints. This presents serious challenges of domain generalisation (DG), domain adaptation (DA), and domain shifting. To enhance the power of ML tools, we can rebuild and retrain models from scratch or we can perform transfer learning. In this paper, we present a comparison study between vision-based technologies for domain-specific and domain-generalised methods. In this research we highlight that simple convolutional neural network (CNN) based deep learning methods perform poorly when they have to tackle domain shifting. Experiments are conducted on two popular vision-based benchmarks, PACS and Office-Home. We introduce an implementation pipeline for domain generalisation methods and conventional deep learning models. The outcome confirms that CNN-based deep learning models show poor generalisation compare to other extensive methods.
As transfer learning techniques are increasingly used to transfer knowledge from the source model to the target task, it becomes important to quantify which source models are suitable for a given target task without performing computationally expensive fine tuning. In this work, we propose HASTE (HArd Subset TransfErability), a new strategy to estimate the transferability of a source model to a particular target task using only a harder subset of target data. By leveraging the internal and output representations of model, we introduce two techniques, one class agnostic and another class specific, to identify harder subsets and show that HASTE can be used with any existing transferability metric to improve their reliability. We further analyze the relation between HASTE and the optimal average log likelihood as well as negative conditional entropy and empirically validate our theoretical bounds. Our experimental results across multiple source model architectures, target datasets, and transfer learning tasks show that HASTE modified metrics are consistently better or on par with the state of the art transferability metrics.
Transfer learning is a promising method for AOI applications since it can significantly shorten sample collection time and improve efficiency in today's smart manufacturing. However, related research enhanced the network models by applying TL without considering the domain similarity among datasets, the data long-tailedness of a source dataset, and mainly used linear transformations to mitigate the lack of samples. This research applies model-based TL via domain similarity to improve the overall performance and data augmentation in both target and source domains to enrich the data quality and reduce the imbalance. Given a group of source datasets from similar industrial processes, we define which group is the most related to the target through the domain discrepancy score and the number of samples each has. Then, we transfer the chosen pre-trained backbone weights to train and fine-tune the target network. Our research suggests increases in the F1 score and the PR curve up to 20% compared with TL using benchmark datasets.
Many self-supervised learning methods are pre-trained on the well-curated ImageNet-1K dataset. In this work, given the excellent scalability of web data, we consider self-supervised pre-training on noisy web sourced image-text paired data. First, we conduct a benchmark study of representative self-supervised pre-training methods on large-scale web data in a like-for-like setting. We compare a range of methods, including single-modal ones that use masked training objectives and multi-modal ones that use image-text constrastive training. We observe that existing multi-modal methods do not outperform their single-modal counterparts on vision transfer learning tasks. We derive an information-theoretical view to explain these benchmark results, which provides insight into how to design a novel vision learner. Inspired by this insight, we present a new visual representation pre-training method, MUlti-modal Generator~(MUG), that learns from scalable web sourced image-text data. MUG achieves state-of-the-art transfer performance on a variety of tasks and demonstrates promising scaling properties. Pre-trained models and code will be made public upon acceptance.
Research on prompting has shown excellent performance with little or even no supervised training across many tasks. However, prompting for machine translation is still under-explored in the literature. We fill this gap by offering a systematic study on prompting strategies for translation, examining various factors for prompt template and demonstration example selection. We further explore the use of monolingual data and the feasibility of cross-lingual, cross-domain, and sentence-to-document transfer learning in prompting. Extensive experiments with GLM-130B (Zeng et al., 2022) as the testbed show that 1) the number and the quality of prompt examples matter, where using suboptimal examples degenerates translation; 2) several features of prompt examples, such as semantic similarity, show significant Spearman correlation with their prompting performance; yet, none of the correlations are strong enough; 3) using pseudo parallel prompt examples constructed from monolingual data via zero-shot prompting could improve translation; and 4) improved performance is achievable by transferring knowledge from prompt examples selected in other settings. We finally provide an analysis on the model outputs and discuss several problems that prompting still suffers from.
In-context learning (ICL) is a type of prompting where a transformer model operates on a sequence of (input, output) examples and performs inference on-the-fly. In this work, we formalize in-context learning as an algorithm learning problem where a transformer model implicitly constructs a hypothesis function at inference-time. We first explore the statistical aspects of this abstraction through the lens of multitask learning: We obtain generalization bounds for ICL when the input prompt is (1) a sequence of i.i.d. (input, label) pairs or (2) a trajectory arising from a dynamical system. The crux of our analysis is relating the excess risk to the stability of the algorithm implemented by the transformer. We characterize when transformer/attention architecture provably obeys the stability condition and also provide empirical verification. For generalization on unseen tasks, we identify an inductive bias phenomenon in which the transfer learning risk is governed by the task complexity and the number of MTL tasks in a highly predictable manner. Finally, we provide numerical evaluations that (1) demonstrate transformers can indeed implement near-optimal algorithms on classical regression problems with i.i.d. and dynamic data, (2) provide insights on stability, and (3) verify our theoretical predictions.
Breast cancer is one of the most prevalent cancers worldwide and pathologists are closely involved in establishing a diagnosis. Tools to assist in making a diagnosis are required to manage the increasing workload. In this context, artificial intelligence (AI) and deep-learning based tools may be used in daily pathology practice. However, it is challenging to develop fast and reliable algorithms that can be trusted by practitioners, whatever the medical center. We describe a patch-based algorithm that incorporates a convolutional neural network to detect and locate invasive carcinoma on breast whole-slide images. The network was trained on a dataset extracted from a reference acquisition center. We then performed a calibration step based on transfer learning to maintain the performance when translating on a new target acquisition center by using a limited amount of additional training data. Performance was evaluated using classical binary measures (accuracy, recall, precision) for both centers (referred to as test reference dataset and test target dataset) and at two levels: patch and slide level. At patch level, accuracy, recall, and precision of the model on the reference and target test sets were 92.1\% and 96.3\%, 95\% and 87.8\%, and 73.9\% and 70.6\%, respectively. At slide level, accuracy, recall, and precision were 97.6\% and 92.0\%, 90.9\% and 100\%, and 100\% and 70.8\% for test sets 1 and 2, respectively. The high performance of the algorithm at both centers shows that the calibration process is efficient. This is performed using limited training data from the new target acquisition center and requires that the model is trained beforehand on a large database from a reference center. This methodology allows the implementation of AI diagnostic tools to help in routine pathology practice.
Rice is the number one staple food in the country, as this serves as the primary livelihood for thousands of Filipino households. However, as the tradition continues, farmers are not familiar with the different types of rice leaf diseases that might compromise the entire rice crop. The need to address the common bacterial leaf blight in rice is a serious disease that can lead to reduced yields and even crop loss of up to 75%. This paper is a design and development of a rice leaf disease detection mobile application prototype using an algorithm used for image analysis. The researchers also used the Rice Disease Image Dataset by Huy Minh Do available at https://www.kaggle.com/ to train state-of-the-art convolutional neural networks using transfer learning. Moreover, we used image augmentation to increase the number of image samples and the accuracy of the neural networks as well
An individualized treatment regime (ITR) is a decision rule that assigns treatments based on patients' characteristics. The value function of an ITR is the expected outcome in a counterfactual world had this ITR been implemented. Recently, there has been increasing interest in combining heterogeneous data sources, such as leveraging the complementary features of randomized controlled trial (RCT) data and a large observational study (OS). Usually, a covariate shift exists between the source and target population, rendering the source-optimal ITR unnecessarily optimal for the target population. We present an efficient and robust transfer learning framework for estimating the optimal ITR with right-censored survival data that generalizes well to the target population. The value function accommodates a broad class of functionals of survival distributions, including survival probabilities and restrictive mean survival times (RMSTs). We propose a doubly robust estimator of the value function, and the optimal ITR is learned by maximizing the value function within a pre-specified class of ITRs. We establish the $N^{-1/3}$ rate of convergence for the estimated parameter indexing the optimal ITR, and show that the proposed optimal value estimator is consistent and asymptotically normal even with flexible machine learning methods for nuisance parameter estimation. We evaluate the empirical performance of the proposed method by simulation studies and a real data application of sodium bicarbonate therapy for patients with severe metabolic acidaemia in the intensive care unit (ICU), combining a RCT and an observational study with heterogeneity.
Standard fine-tuning of language models typically performs well on in-distribution data, but suffers with generalization to distribution shifts. In this work, we aim to improve the generalization of adapter-based cross-lingual task transfer where such cross-language distribution shifts are imminent. We investigate scheduled unfreezing algorithms -- originally proposed to mitigate catastrophic forgetting in transfer learning -- for fine-tuning task adapters. Our experiments show that scheduled unfreezing methods close the gap to full fine-tuning and achieve stronger cross-lingual transfer performance, suggesting that these methods can go beyond just mitigating catastrophic forgetting. Next, aiming to understand these empirical findings, we investigate the learning dynamics of scheduled unfreezing using Fisher Information. Our experiments reveal that scheduled unfreezing induces different learning dynamics compared to standard fine-tuning, and provide evidence that the dynamics of Fisher Information during training correlate with cross-lingual generalization performance. We additionally propose a general scheduled unfreezing algorithm that achieves an average of 2 points improvement over four datasets compared to standard fine-tuning and provides empirical evidence for a theory-based justification of the heuristic unfreezing schedule for adapter training.
Coordination is one of the most difficult aspects of multi-agent reinforcement learning (MARL). One reason is that agents normally choose their actions independently of one another. In order to see coordination strategies emerging from the combination of independent policies, the recent research has focused on the use of a centralized function (CF) that learns each agent's contribution to the team reward. However, the structure in which the environment is presented to the agents and to the CF is typically overlooked. We have observed that the features used to describe the coordination problem can be represented as vertex features of a latent graph structure. Here, we present TransfQMix, a new approach that uses transformers to leverage this latent structure and learn better coordination policies. Our transformer agents perform a graph reasoning over the state of the observable entities. Our transformer Q-mixer learns a monotonic mixing-function from a larger graph that includes the internal and external states of the agents. TransfQMix is designed to be entirely transferable, meaning that same parameters can be used to control and train larger or smaller teams of agents. This enables to deploy promising approaches to save training time and derive general policies in MARL, such as transfer learning, zero-shot transfer, and curriculum learning. We report TransfQMix's performances in the Spread and StarCraft II environments. In both settings, it outperforms state-of-the-art Q-Learning models, and it demonstrates effectiveness in solving problems that other methods can not solve.