Language identification describes the task of recognizing the language of written text in documents. This information is crucial because it can be used to support the analysis of a document's vocabulary and context. Supervised learning methods in recent years have advanced the task of language identification. However, these methods usually require large labeled datasets, which often need to be included for various domains of images, such as documents or scene images. In this work, we propose DocLangID, a transfer learning approach to identify the language of unlabeled historical documents. We achieve this by first leveraging labeled data from a different but related domain of historical documents. Secondly, we implement a distance-based few-shot learning approach to adapt a convolutional neural network to new languages of the unlabeled dataset. By introducing small amounts of manually labeled examples from the set of unlabeled images, our feature extractor develops a better adaptability towards new and different data distributions of historical documents. We show that such a model can be effectively fine-tuned for the unlabeled set of images by only reusing the same few-shot examples. We showcase our work across 10 languages that mostly use the Latin script. Our experiments on historical documents demonstrate that our combined approach improves the language identification performance, achieving 74% recognition accuracy on the four unseen languages of the unlabeled dataset.
Contrastive learning has become a popular approach in natural language processing, particularly for the learning of sentence embeddings. However, the discrete nature of natural language makes it difficult to ensure the quality of positive and negative sample pairs generated through data augmentation methods. Although supervised contrastive learning can produce more accurate sample pairs with human feedback labels, it still lacks fine-grained training signals. In this paper, we propose to improve \textbf{C}ontrastive \textbf{L}earning of sentence embeddings from \textbf{AI} \textbf{F}eedback \textbf{(CLAIF)}. Our method utilizes AI feedback from large pre-trained language models (LLMs) to construct sample pairs with fine-grained sample similarity scores to improve contrastive learning. Besides, we combine human feedback and AI feedback to provide better supervision signals for supervised contrastive learning of sentence embeddings. Experimental results show that our method achieves state-of-the-art performance on several semantic textual similarity (STS) and transfer learning tasks compared to other unsupervised and supervised contrastive learning methods.
In this era of information explosion, deceivers use different domains or mediums of information to exploit the users, such as News, Emails, and Tweets. Although numerous research has been done to detect deception in all these domains, information shortage in a new event necessitates these domains to associate with each other to battle deception. To form this association, we propose a feature augmentation method by harnessing the intermediate layer representation of neural models. Our approaches provide an improvement over the self-domain baseline models by up to 6.60%. We find Tweets to be the most helpful information provider for Fake News and Phishing Email detection, whereas News helps most in Tweet Rumor detection. Our analysis provides a useful insight for domain knowledge transfer which can help build a stronger deception detection system than the existing literature.
Existing semantic segmentation approaches are often limited by costly pixel-wise annotations and predefined classes. In this work, we present CLIP-S$^4$ that leverages self-supervised pixel representation learning and vision-language models to enable various semantic segmentation tasks (e.g., unsupervised, transfer learning, language-driven segmentation) without any human annotations and unknown class information. We first learn pixel embeddings with pixel-segment contrastive learning from different augmented views of images. To further improve the pixel embeddings and enable language-driven semantic segmentation, we design two types of consistency guided by vision-language models: 1) embedding consistency, aligning our pixel embeddings to the joint feature space of a pre-trained vision-language model, CLIP; and 2) semantic consistency, forcing our model to make the same predictions as CLIP over a set of carefully designed target classes with both known and unknown prototypes. Thus, CLIP-S$^4$ enables a new task of class-free semantic segmentation where no unknown class information is needed during training. As a result, our approach shows consistent and substantial performance improvement over four popular benchmarks compared with the state-of-the-art unsupervised and language-driven semantic segmentation methods. More importantly, our method outperforms these methods on unknown class recognition by a large margin.
Big data has become the norm in astronomy, making it an ideal domain for computer science research. Astronomers typically classify galaxies based on their morphologies, a practice that dates back to Hubble (1936). With small datasets, classification could be performed by individuals or small teams, but the exponential growth of data from modern telescopes necessitates automated classification methods. In December 2013, Winton Capital, Galaxy Zoo, and the Kaggle team created the Galaxy Challenge, which tasked participants with developing models to classify galaxies. The Kaggle Galaxy Zoo dataset has since been widely used by researchers. This study investigates the impact of colour space transformation on classification accuracy and explores the effect of CNN architecture on this relationship. Multiple colour spaces (RGB, XYZ, LAB, etc.) and CNN architectures (VGG, ResNet, DenseNet, Xception, etc.) are considered, utilizing pre-trained models and weights. However, as most pre-trained models are designed for natural RGB images, we examine their performance with transformed, non-natural astronomical images. We test our hypothesis by evaluating individual networks with RGB and transformed colour spaces and examining various ensemble configurations. A minimal hyperparameter search ensures optimal results. Our findings indicate that using transformed colour spaces in individual networks yields higher validation accuracy, and ensembles of networks and colour spaces further improve accuracy. This research aims to validate the utility of colour space transformation for astronomical image classification and serve as a benchmark for future studies.
Recently, self-supervised learning (SSL) from unlabelled speech data has gained increased attention in the automatic speech recognition (ASR) community. Typical SSL methods include autoregressive predictive coding (APC), Wav2vec2.0, and hidden unit BERT (HuBERT). However, SSL models are biased to the pretraining data. When SSL models are finetuned with data from another domain, domain shifting occurs and might cause limited knowledge transfer for downstream tasks. In this paper, we propose a novel framework, domain responsible adaptation and finetuning (DRAFT), to reduce domain shifting in pretrained speech models, and evaluate it for a causal and non-causal transformer. For the causal transformer, an extension of APC (E-APC) is proposed to learn richer information from unlabelled data by using multiple temporally-shifted sequences to perform prediction. For the non-causal transformer, various solutions for using the bidirectional APC (Bi-APC) are investigated. In addition, the DRAFT framework is examined for Wav2vec2.0 and HuBERT methods, which use non-causal transformers as the backbone. The experiments are conducted on child ASR (using the OGI and MyST databases) using SSL models trained with unlabelled adult speech data from Librispeech. The relative WER improvements of up to 19.7% on the two child tasks are observed when compared to the pretrained models without adaptation. With the proposed methods (E-APC and DRAFT), the relative WER improvements are even larger (30% and 19% on the OGI and MyST data, respectively) when compared to the models without using pretraining methods.
This paper describes our system developed for the SemEval-2023 Task 12 "Sentiment Analysis for Low-resource African Languages using Twitter Dataset". Sentiment analysis is one of the most widely studied applications in natural language processing. However, most prior work still focuses on a small number of high-resource languages. Building reliable sentiment analysis systems for low-resource languages remains challenging, due to the limited training data in this task. In this work, we propose to leverage language-adaptive and task-adaptive pretraining on African texts and study transfer learning with source language selection on top of an African language-centric pretrained language model. Our key findings are: (1) Adapting the pretrained model to the target language and task using a small yet relevant corpus improves performance remarkably by more than 10 F1 score points. (2) Selecting source languages with positive transfer gains during training can avoid harmful interference from dissimilar languages, leading to better results in multilingual and cross-lingual settings. In the shared task, our system wins 8 out of 15 tracks and, in particular, performs best in the multilingual evaluation.
We present the findings of our participation in the SemEval-2023 Task 10: Explainable Detection of Online Sexism (EDOS) task, a shared task on offensive language (sexism) detection on English Gab and Reddit dataset. We investigated the effects of transferring two language models: XLM-T (sentiment classification) and HateBERT (same domain -- Reddit) for multi-level classification into Sexist or not Sexist, and other subsequent sub-classifications of the sexist data. We also use synthetic classification of unlabelled dataset and intermediary class information to maximize the performance of our models. We submitted a system in Task A, and it ranked 49th with F1-score of 0.82. This result showed to be competitive as it only under-performed the best system by 0.052% F1-score.
Modern machine learning increasingly supports paradigms that are multi-institutional (using data from multiple institutions during training) or cross-institutional (using models from multiple institutions for inference), but the empirical effects of these paradigms are not well understood. This study investigates cross-institutional learning via an empirical case study in higher education. We propose a framework and metrics for assessing the utility and fairness of student dropout prediction models that are transferred across institutions. We examine the feasibility of cross-institutional transfer under real-world data- and model-sharing constraints, quantifying model biases for intersectional student identities, characterizing potential disparate impact due to these biases, and investigating the impact of various cross-institutional ensembling approaches on fairness and overall model performance. We perform this analysis on data representing over 200,000 enrolled students annually from four universities without sharing training data between institutions. We find that a simple zero-shot cross-institutional transfer procedure can achieve similar performance to locally-trained models for all institutions in our study, without sacrificing model fairness. We also find that stacked ensembling provides no additional benefits to overall performance or fairness compared to either a local model or the zero-shot transfer procedure we tested. We find no evidence of a fairness-accuracy tradeoff across dozens of models and transfer schemes evaluated. Our auditing procedure also highlights the importance of intersectional fairness analysis, revealing performance disparities at the intersection of sensitive identity groups that are concealed under one-dimensional analysis.
Federated Semi-supervised Learning (FedSSL) has emerged as a new paradigm for allowing distributed clients to collaboratively train a machine learning model over scarce labeled data and abundant unlabeled data. However, existing works for FedSSL rely on a closed-world assumption that all local training data and global testing data are from seen classes observed in the labeled dataset. It is crucial to go one step further: adapting FL models to an open-world setting, where unseen classes exist in the unlabeled data. In this paper, we propose a novel Federatedopen-world Semi-Supervised Learning (FedoSSL) framework, which can solve the key challenge in distributed and open-world settings, i.e., the biased training process for heterogeneously distributed unseen classes. Specifically, since the advent of a certain unseen class depends on a client basis, the locally unseen classes (exist in multiple clients) are likely to receive differentiated superior aggregation effects than the globally unseen classes (exist only in one client). We adopt an uncertainty-aware suppressed loss to alleviate the biased training between locally unseen and globally unseen classes. Besides, we enable a calibration module supplementary to the global aggregation to avoid potential conflicting knowledge transfer caused by inconsistent data distribution among different clients. The proposed FedoSSL can be easily adapted to state-of-the-art FL methods, which is also validated via extensive experiments on benchmarks and real-world datasets (CIFAR-10, CIFAR-100 and CINIC-10).
Non-IID transfer learning on graphs is crucial in many high-stakes domains. The majority of existing works assume stationary distribution for both source and target domains. However, real-world graphs are intrinsically dynamic, presenting challenges in terms of domain evolution and dynamic discrepancy between source and target domains. To bridge the gap, we shift the problem to the dynamic setting and pose the question: given the label-rich source graphs and the label-scarce target graphs both observed in previous T timestamps, how can we effectively characterize the evolving domain discrepancy and optimize the generalization performance of the target domain at the incoming T+1 timestamp? To answer it, we propose a generalization bound for dynamic non-IID transfer learning on graphs, which implies the generalization performance is dominated by domain evolution and domain discrepancy between source and target graphs. Inspired by the theoretical results, we introduce a novel generic framework named EvoluNet. It leverages a transformer-based temporal encoding module to model temporal information of the evolving domains and then uses a dynamic domain unification module to efficiently learn domain-invariant representations across the source and target domains. Finally, EvoluNet outperforms the state-of-the-art models by up to 12.1%, demonstrating its effectiveness in transferring knowledge from dynamic source graphs to dynamic target graphs.
Automatic music transcription (AMT) is one of the most challenging tasks in the music information retrieval domain. It is the process of converting an audio recording of music into a symbolic representation containing information about the notes, chords, and rhythm. Current research in this domain focuses on developing new models based on transformer architecture or using methods to perform semi-supervised training, which gives outstanding results, but the computational cost of training such models is enormous. This work shows how to employ easily generated synthesized audio data produced by software synthesizers to train a universal model. It is a good base for further transfer learning to quickly adapt transcription model for other instruments. Achieved results prove that using synthesized data for training may be a good base for pretraining general-purpose models, where the task of transcription is not focused on one instrument.
Machine Learning (ML) is of increasing interest for modeling parametric effects in manufacturing processes. But this approach is limited to established processes for which a deep physics-based understanding has been developed over time, since state-of-the-art approaches focus on reducing the experimental and/or computational costs of generating the training data but ignore the inherent and significant cost of developing qualitatively accurate physics-based models for new processes . This paper proposes a transfer learning based approach to address this issue, in which a ML model is trained on a large amount of computationally inexpensive data from a physics-based process model (source) and then fine-tuned on a smaller amount of costly experimental data (target). The novelty lies in pushing the boundaries of the qualitative accuracy demanded of the source model, which is assumed to be high in the literature, and is the root of the high model development cost. Our approach is evaluated for modeling the printed line width in Fused Filament Fabrication. Despite extreme functional and quantitative inaccuracies in the source our approach reduces the model development cost by years, experimental cost by 56-76%, computational cost by orders of magnitude, and prediction error by 16-24%.
Theoretical studies on transfer learning or domain adaptation have so far focused on situations with a known hypothesis class or model; however in practice, some amount of model selection is usually involved, often appearing under the umbrella term of hyperparameter-tuning: for example, one may think of the problem of tuning for the right neural network architecture towards a target task, while leveraging data from a related source task. Now, in addition to the usual tradeoffs on approximation vs estimation errors involved in model selection, this problem brings in a new complexity term, namely, the transfer distance between source and target distributions, which is known to vary with the choice of hypothesis class. We present a first study of this problem, focusing on classification; in particular, the analysis reveals some remarkable phenomena: adaptive rates, i.e., those achievable with no distributional information, can be arbitrarily slower than oracle rates, i.e., when given knowledge on distances.
Transfer learning has been widely adopted for few-shot classification. Recent studies reveal that obtaining good generalization representation of images on novel classes is the key to improving the few-shot classification accuracy. To address this need, we prove theoretically that leveraging ensemble learning on the base classes can correspondingly reduce the true error in the novel classes. Following this principle, a novel method named Ensemble Learning with Multi-Order Statistics (ELMOS) is proposed in this paper. In this method, after the backbone network, we use multiple branches to create the individual learners in the ensemble learning, with the goal to reduce the storage cost. We then introduce different order statistics pooling in each branch to increase the diversity of the individual learners. The learners are optimized with supervised losses during the pre-training phase. After pre-training, features from different branches are concatenated for classifier evaluation. Extensive experiments demonstrate that each branch can complement the others and our method can produce a state-of-the-art performance on multiple few-shot classification benchmark datasets.
Colon polyps are considered important precursors for colorectal cancer. Automatic segmentation of colon polyps can significantly reduce the misdiagnosis of colon cancer and improve physician annotation efficiency. While many methods have been proposed for polyp segmentation, training large-scale segmentation networks with limited colonoscopy data remains a challenge. Recently, the Segment Anything Model (SAM) has recently gained much attention in both natural and medical image segmentation. SAM demonstrates superior performance in several image benchmarks and therefore shows great potential for medical image segmentation. In this study, we propose Poly-SAM, a finetuned SAM model for polyp segmentation, and compare its performance to several state-of-the-art polyp segmentation models. We also compare two transfer learning strategies of SAM with and without finetuning its encoders. Evaluated on five public datasets, our Polyp-SAM achieves state-of-the-art performance on two datasets and impressive performance on three datasets, with dice scores all above 88%. This study demonstrates the great potential of adapting SAM to medical image segmentation tasks. We plan to release the code and model weights for this paper at: https://github.com/ricklisz/Polyp-SAM.
Globally, Coronary Heart Disease (CHD) is one of the main causes of death. Early detection of CHD can improve patient outcomes and reduce mortality rates. We propose a novel framework for predicting the presence of CHD using a combination of machine learning and image processing techniques. The framework comprises various phases, including analyzing the data, feature selection using ReliefF, 3D CNN-based segmentation, feature extraction by means of transfer learning, feature fusion as well as classification, and Adagrad optimization. The first step of the proposed framework involves analyzing the data to identify patterns and correlations that may be indicative of CHD. Next, ReliefF feature selection is applied to decide on the most relevant features from the sample images. The 3D CNN-based segmentation technique is then used to segment the optic disc and macula, which are important regions for CHD diagnosis. Feature extraction using transfer learning is performed to extract features from the segmented regions of interest. The extracted features are then fused using a feature fusion technique, and a classifier is trained to predict the presence of CHD. Finally, Adagrad optimization is used to optimize the performance of the classifier. Our framework is evaluated on a dataset of sample images collected from patients with and without CHD. The results show that the anticipated framework accomplishes elevated accuracy in predicting the presence of CHD. either a particular user with a reasonable degree of accuracy compared to the previously employed classifiers like SVM, etc.
Energy optimization leveraging artificially intelligent algorithms has been proven effective. However, when buildings are commissioned, there is no historical data that could be used to train these algorithms. On-line Reinforcement Learning (RL) algorithms have shown significant promise, but their deployment carries a significant risk, because as the RL agent initially explores its action space it could cause significant discomfort to the building residents. In this paper we present ReLBOT - a new technique that uses transfer learning in conjunction with deep RL to transfer knowledge from an existing, optimized and instrumented building, to the newly commissioning smart building, to reduce the adverse impact of the reinforcement learning agent's warm-up period. We demonstrate improvements of up to 6.2 times in the duration, and up to 132 times in prediction variance, for the reinforcement learning agent's warm-up period.
We introduce the concept of Procedural Content Generation via Knowledge Transformation (PCG-KT), a new lens and framework for characterizing PCG methods and approaches in which content generation is enabled by the process of knowledge transformation -- transforming knowledge derived from one domain in order to apply it in another. Our work is motivated by a substantial number of recent PCG works that focus on generating novel content via repurposing derived knowledge. Such works have involved, for example, performing transfer learning on models trained on one game's content to adapt to another game's content, as well as recombining different generative distributions to blend the content of two or more games. Such approaches arose in part due to limitations in PCG via Machine Learning (PCGML) such as producing generative models for games lacking training data and generating content for entirely new games. In this paper, we categorize such approaches under this new lens of PCG-KT by offering a definition and framework for describing such methods and surveying existing works using this framework. Finally, we conclude by highlighting open problems and directions for future research in this area.
Code-trained language models have proven to be highly effective for various code intelligence tasks. However, they can be challenging to train and deploy for many software engineering applications due to computational bottlenecks and memory constraints. Implementing effective strategies to address these issues requires a better understanding of these 'black box' models. In this paper, we perform the first neuron-level analysis for source code models to identify \textit{important} neurons within latent representations. We achieve this by eliminating neurons that are highly similar or irrelevant to the given task. This approach helps us understand which neurons and layers can be eliminated (redundancy analysis) and where important code properties are located within the network (concept analysis). Using redundancy analysis, we make observations relevant to knowledge transfer and model optimization applications. We find that over 95\% of the neurons are redundant with respect to our code intelligence tasks and can be eliminated without significant loss in accuracy. We also discover several subsets of neurons that can make predictions with baseline accuracy. Through concept analysis, we explore the traceability and distribution of human-recognizable concepts within latent code representations which could be used to influence model predictions. We trace individual and subsets of important neurons to specific code properties and identify 'number' neurons, 'string' neurons, and higher-level 'text' neurons for token-level tasks and higher-level concepts important for sentence-level downstream tasks. This also helps us understand how decomposable and transferable task-related features are and can help devise better techniques for transfer learning, model compression, and the decomposition of deep neural networks into modules.