Data for pretraining machine learning models often consists of collections of heterogeneous datasets. Although training on their union is reasonable in agnostic settings, it might be suboptimal when the target domain -- where the model will ultimately be used -- is known in advance. In that case, one would ideally pretrain only on the dataset(s) most similar to the target one. Instead of limiting this choice to those datasets already present in the pretraining collection, here we explore extending this search to all datasets that can be synthesized as `combinations' of them. We define such combinations as multi-dataset interpolations, formalized through the notion of generalized geodesics from optimal transport (OT) theory. We compute these geodesics using a recent notion of distance between labeled datasets, and derive alternative interpolation schemes based on it: using either barycentric projections or optimal transport maps, the latter computed using recent neural OT methods. These methods are scalable, efficient, and -- notably -- can be used to interpolate even between datasets with distinct and unrelated label sets. Through various experiments in transfer learning in computer vision, we demonstrate this is a promising new approach for targeted on-demand dataset synthesis.
With the increasing utilization of Internet of Things (IoT) enabled drones in diverse applications like photography, delivery, and surveillance, concerns regarding privacy and security have become more prominent. Drones have the ability to capture sensitive information, compromise privacy, and pose security risks. As a result, the demand for advanced technology to automate drone detection has become crucial. This paper presents a project on a transfer-based drone detection scheme, which forms an integral part of a computer vision-based module and leverages transfer learning to enhance performance. By harnessing the knowledge of pre-trained models from a related domain, transfer learning enables improved results even with limited training data. To evaluate the scheme's performance, we conducted tests on benchmark datasets, including the Drone-vs-Bird Dataset and the UAVDT dataset. Notably, the scheme's effectiveness is highlighted by its IOU-based validation results, demonstrating the potential of deep learning-based technology in automating drone detection in critical areas such as airports, military bases, and other high-security zones.
With rapid progress in deep learning, neural networks have been widely used in scientific research and engineering applications as surrogate models. Despite the great success of neural networks in fitting complex systems, two major challenges still remain: i) the lack of generalization on different problems/datasets, and ii) the demand for large amounts of simulation data that are computationally expensive. To resolve these challenges, we propose the differentiable \mf (DMF) model, which leverages neural architecture search (NAS) to automatically search the suitable model architecture for different problems, and transfer learning to transfer the learned knowledge from low-fidelity (fast but inaccurate) data to high-fidelity (slow but accurate) model. Novel and latest machine learning techniques such as hyperparameters search and alternate learning are used to improve the efficiency and robustness of DMF. As a result, DMF can efficiently learn the physics simulations with only a few high-fidelity training samples, and outperform the state-of-the-art methods with a significant margin (with up to 58$\%$ improvement in RMSE) based on a variety of synthetic and practical benchmark problems.
Instruction tuning of language models has demonstrated the ability to enhance model generalization to unseen tasks via in-context learning using a few examples. However, typical supervised learning still requires a plethora of downstream training data for finetuning. Often in real-world situations, there is a scarcity of data available for finetuning, falling somewhere between few shot inference and fully supervised finetuning. In this work, we demonstrate the sample efficiency of instruction tuned models over various tasks by estimating the minimal downstream training data required by them to perform transfer learning and match the performance of state-of-the-art (SOTA) supervised models. We conduct experiments on 119 tasks from Super Natural Instructions (SuperNI) in both the single task learning (STL) and multi task learning (MTL) settings. Our findings reveal that, in the STL setting, instruction tuned models equipped with 25% of the downstream train data surpass the SOTA performance on the downstream tasks. In the MTL setting, an instruction tuned model trained on only 6% of downstream training data achieve SOTA, while using 100% of the training data results in a 3.69% points improvement (ROUGE-L 74.68) over the previous SOTA. We conduct an analysis on T5 vs Tk-Instruct by developing several baselines to demonstrate that instruction tuning aids in increasing both sample efficiency and transfer learning. Additionally, we observe a consistent ~4% performance increase in both settings when pre-finetuning is performed with instructions. Finally, we conduct a categorical study and find that contrary to previous results, tasks in the question rewriting and title generation categories suffer from instruction tuning.
Neural network (NN) compression via techniques such as pruning, quantization requires setting compression hyperparameters (e.g., number of channels to be pruned, bitwidths for quantization) for each layer either manually or via neural architecture search (NAS) which can be computationally expensive. We address this problem by providing an end-to-end technique that optimizes for model's Floating Point Operations (FLOPs) or for on-device latency via a novel $\frac{\ell_1}{\ell_2}$ latency surrogate. Our algorithm is versatile and can be used with many popular compression methods including pruning, low-rank factorization, and quantization. Crucially, it is fast and runs in almost the same amount of time as single model training; which is a significant training speed-up over standard NAS methods. For BERT compression on GLUE fine-tuning tasks, we achieve $50\%$ reduction in FLOPs with only $1\%$ drop in performance. For compressing MobileNetV3 on ImageNet-1K, we achieve $15\%$ reduction in FLOPs, and $11\%$ reduction in on-device latency without drop in accuracy, while still requiring $3\times$ less training compute than SOTA compression techniques. Finally, for transfer learning on smaller datasets, our technique identifies $1.2\times$-$1.4\times$ cheaper architectures than standard MobileNetV3, EfficientNet suite of architectures at almost the same training cost and accuracy.
The detection of emotions using an Electroencephalogram (EEG) is a crucial area in brain-computer interfaces and has valuable applications in fields such as rehabilitation and medicine. In this study, we employed transfer learning to overcome the challenge of limited data availability in EEG-based emotion detection. The base model used in this study was Resnet50. Additionally, we employed a novel feature combination in EEG-based emotion detection. The input to the model was in the form of an image matrix, which comprised Mean Phase Coherence (MPC) and Magnitude Squared Coherence (MSC) in the upper-triangular and lower-triangular matrices, respectively. We further improved the technique by incorporating features obtained from the Differential Entropy (DE) into the diagonal, which previously held little to no useful information for classifying emotions. The dataset used in this study, SEED EEG (62 channel EEG), comprises three classes (Positive, Neutral, and Negative). We calculated both subject-independent and subject-dependent accuracy. The subject-dependent accuracy was obtained using a 10-fold cross-validation method and was 93.1%, while the subject-independent classification was performed by employing the leave-one-subject-out (LOSO) strategy. The accuracy obtained in subject-independent classification was 71.6%. Both of these accuracies are at least twice better than the chance accuracy of classifying 3 classes. The study found the use of MSC and MPC in EEG-based emotion detection promising for emotion classification. The future scope of this work includes the use of data augmentation techniques, enhanced classifiers, and better features for emotion classification.
Speech-to-text (S2T) summarization is a time-saving technique for filtering and keeping up with the broadcast news uploaded online on a daily basis. The rise of large language models from deep learning with impressive text generation capabilities has placed the research focus on summarization systems that produce paraphrased compact versions of the document content, also known as abstractive summaries. End-to-end (E2E) modelling of S2T abstractive summarization is a promising approach that offers the possibility of generating rich latent representations that leverage non-verbal and acoustic information, as opposed to the use of only linguistic information from automatically generated transcripts in cascade systems. However, the few literature on E2E modelling of this task fails on exploring different domains, namely broadcast news, which is challenging domain where large and diversified volumes of data are presented to the user every day. We model S2T summarization both with a cascade and an E2E system for a corpus of broadcast news in French. Our novel E2E model leverages external data by resorting to transfer learning from a pre-trained T2T summarizer. Experiments show that both our cascade and E2E abstractive summarizers are stronger than an extractive baseline. However, the performance of the E2E model still lies behind the cascade one, which is object of an extensive analysis that includes future directions to close that gap.
Medical caption prediction which can be regarded as a task of medical report generation (MRG), requires the automatic generation of coherent and accurate captions for the given medical images. However, the scarcity of labelled medical image-report pairs presents great challenges in the development of deep and large-scale neural networks capable of harnessing the potential artificial general intelligence power like large language models (LLMs). In this work, we propose customizing off-the-shelf general-purpose large-scale pre-trained models, i.e., foundation models (FMs), in computer vision and natural language processing with a specific focus on medical report generation. Specifically, following BLIP-2, a state-of-the-art vision-language pre-training approach, we introduce our encoder-decoder-based MRG model. This model utilizes a lightweight query Transformer to connect two FMs: the giant vision Transformer EVA-ViT-g and a bilingual LLM trained to align with human intentions (referred to as ChatGLM-6B). Furthermore, we conduct ablative experiments on the trainable components of the model to identify the crucial factors for effective transfer learning. Our findings demonstrate that unfreezing EVA-ViT-g to learn medical image representations, followed by parameter-efficient training of ChatGLM-6B to capture the writing styles of medical reports, is essential for achieving optimal results. Our best attempt (PCLmed Team) achieved the 4th and the 2nd, respectively, out of 13 participating teams, based on the BERTScore and ROUGE-1 metrics, in the ImageCLEFmedical Caption 2023 Caption Prediction Task competition.
Cross-lingual text classification leverages text classifiers trained in a high-resource language to perform text classification in other languages with no or minimal fine-tuning (zero/few-shots cross-lingual transfer). Nowadays, cross-lingual text classifiers are typically built on large-scale, multilingual language models (LMs) pretrained on a variety of languages of interest. However, the performance of these models vary significantly across languages and classification tasks, suggesting that the superposition of the language modelling and classification tasks is not always effective. For this reason, in this paper we propose revisiting the classic "translate-and-test" pipeline to neatly separate the translation and classification stages. The proposed approach couples 1) a neural machine translator translating from the targeted language to a high-resource language, with 2) a text classifier trained in the high-resource language, but the neural machine translator generates "soft" translations to permit end-to-end backpropagation during fine-tuning of the pipeline. Extensive experiments have been carried out over three cross-lingual text classification datasets (XNLI, MLDoc and MultiEURLEX), with the results showing that the proposed approach has significantly improved performance over a competitive baseline.
A challenge in the Dialogue State Tracking (DST) field is adapting models to new domains without using any supervised data, zero-shot domain adaptation. Parameter-Efficient Transfer Learning (PETL) has the potential to address this problem due to its robustness. However, it has yet to be applied to the zero-shot scenarios, as it is not clear how to apply it unsupervisedly. Our method, Prompter, uses descriptions of target domain slots to generate dynamic prefixes that are concatenated to the key and values at each layer's self-attention mechanism. This allows for the use of prefix-tuning in zero-shot. Prompter outperforms previous methods on both the MultiWOZ and SGD benchmarks. In generating prefixes, our analyses find that Prompter not only utilizes the semantics of slot descriptions but also how often the slots appear together in conversation. Moreover, Prompter's gains are due to its improved ability to distinguish "none"-valued dialogue slots, compared against baselines.
Transfer learning is a useful technique for achieving improved performance and reducing training costs by leveraging the knowledge gained from source tasks and applying it to target tasks. Assessing the effectiveness of transfer learning relies on understanding the similarity between the ground truth of the source and target tasks. In real-world applications, tasks often exhibit partial similarity, where certain aspects are similar while others are different or irrelevant. To investigate the impact of partial similarity on transfer learning performance, we focus on a linear regression model with two distinct sets of features: a common part shared across tasks and a task-specific part. Our study explores various types of transfer learning, encompassing two options for parameter transfer. By establishing a theoretical characterization on the error of the learned model, we compare these transfer learning options, particularly examining how generalization performance changes with the number of features/parameters in both underparameterized and overparameterized regimes. Furthermore, we provide practical guidelines for determining the number of features in the common and task-specific parts for improved generalization performance. For example, when the total number of features in the source task's learning model is fixed, we show that it is more advantageous to allocate a greater number of redundant features to the task-specific part rather than the common part. Moreover, in specific scenarios, particularly those characterized by high noise levels and small true parameters, sacrificing certain true features in the common part in favor of employing more redundant features in the task-specific part can yield notable benefits.
The integration of machine learning in medical image analysis can greatly enhance the quality of healthcare provided by physicians. The combination of human expertise and computerized systems can result in improved diagnostic accuracy. An automated machine learning approach simplifies the creation of custom image recognition models by utilizing neural architecture search and transfer learning techniques. Medical imaging techniques are used to non-invasively create images of internal organs and body parts for diagnostic and procedural purposes. This article aims to highlight the potential applications, strategies, and techniques of AutoML in medical imaging through theoretical and empirical evidence.
Cross-Lingual Semantic Parsing (CLSP) aims to translate queries in multiple natural languages (NLs) into meaning representations (MRs) such as SQL, lambda calculus, and logic forms. However, existing CLSP models are separately proposed and evaluated on datasets of limited tasks and applications, impeding a comprehensive and unified evaluation of CLSP on a diverse range of NLs and MRs. To this end, we present XSemPLR, a unified benchmark for cross-lingual semantic parsing featured with 22 natural languages and 8 meaning representations by examining and selecting 9 existing datasets to cover 5 tasks and 164 domains. We use XSemPLR to conduct a comprehensive benchmark study on a wide range of multilingual language models including encoder-based models (mBERT, XLM-R), encoder-decoder models (mBART, mT5), and decoder-based models (Codex, BLOOM). We design 6 experiment settings covering various lingual combinations (monolingual, multilingual, cross-lingual) and numbers of learning samples (full dataset, few-shot, and zero-shot). Our experiments show that encoder-decoder models (mT5) achieve the highest performance compared with other popular models, and multilingual training can further improve the average performance. Notably, multilingual large language models (e.g., BLOOM) are still inadequate to perform CLSP tasks. We also find that the performance gap between monolingual training and cross-lingual transfer learning is still significant for multilingual models, though it can be mitigated by cross-lingual few-shot training. Our dataset and code are available at https://github.com/psunlpgroup/XSemPLR.
This work introduces Zambezi Voice, an open-source multilingual speech resource for Zambian languages. It contains two collections of datasets: unlabelled audio recordings of radio news and talk shows programs (160 hours) and labelled data (over 80 hours) consisting of read speech recorded from text sourced from publicly available literature books. The dataset is created for speech recognition but can be extended to multilingual speech processing research for both supervised and unsupervised learning approaches. To our knowledge, this is the first multilingual speech dataset created for Zambian languages. We exploit pretraining and cross-lingual transfer learning by finetuning the Wav2Vec2.0 large-scale multilingual pre-trained model to build end-to-end (E2E) speech recognition models for our baseline models. The dataset is released publicly under a Creative Commons BY-NC-ND 4.0 license and can be accessed via https://github.com/unza-speech-lab/zambezi-voice .
In this paper, we are comparing several methods of training the Slovak speech recognition models based on the Transformers architecture. Specifically, we are exploring the approach of transfer learning from the existing Czech pre-trained Wav2Vec 2.0 model into Slovak. We are demonstrating the benefits of the proposed approach on three Slovak datasets. Our Slovak models scored the best results when initializing the weights from the Czech model at the beginning of the pre-training phase. Our results show that the knowledge stored in the Cezch pre-trained model can be successfully reused to solve tasks in Slovak while outperforming even much larger public multilingual models.
End-to-end speech summarization (E2E SSum) directly summarizes input speech into easy-to-read short sentences with a single model. This approach is promising because it, in contrast to the conventional cascade approach, can utilize full acoustical information and mitigate to the propagation of transcription errors. However, due to the high cost of collecting speech-summary pairs, an E2E SSum model tends to suffer from training data scarcity and output unnatural sentences. To overcome this drawback, we propose for the first time to integrate a pre-trained language model (LM), which is highly capable of generating natural sentences, into the E2E SSum decoder via transfer learning. In addition, to reduce the gap between the independently pre-trained encoder and decoder, we also propose to transfer the baseline E2E SSum encoder instead of the commonly used automatic speech recognition encoder. Experimental results show that the proposed model outperforms baseline and data augmented models.
Arabic dialect identification (ADI) tools are an important part of the large-scale data collection pipelines necessary for training speech recognition models. As these pipelines require application of ADI tools to potentially out-of-domain data, we aim to investigate how vulnerable the tools may be to this domain shift. With self-supervised learning (SSL) models as a starting point, we evaluate transfer learning and direct classification from SSL features. We undertake our evaluation under rich conditions, with a goal to develop ADI systems from pretrained models and ultimately evaluate performance on newly collected data. In order to understand what factors contribute to model decisions, we carry out a careful human study of a subset of our data. Our analysis confirms that domain shift is a major challenge for ADI models. We also find that while self-training does alleviate this challenges, it may be insufficient for realistic conditions.
Variational quantum circuits (VQCs) hold promise for quantum machine learning on noisy intermediate-scale quantum (NISQ) devices. While tensor-train networks (TTNs) can enhance VQC representation and generalization, the resulting hybrid model, TTN-VQC, faces optimization challenges due to the Polyak-Lojasiewicz (PL) condition. To mitigate this challenge, we introduce Pre+TTN-VQC, a pre-trained TTN model combined with a VQC. Our theoretical analysis, grounded in two-stage empirical risk minimization, provides an upper bound on the transfer learning risk. It demonstrates the approach's advantages in overcoming the optimization challenge while maintaining TTN-VQC's generalization capability. We validate our findings through experiments on quantum dot and handwritten digit classification using simulated and actual NISQ environments.
For prediction and real-time control tasks, machine-learning (ML)-based digital twins are frequently employed. However, while these models are typically accurate, they are custom-designed for individual systems, making system-to-system (S2S) transferability difficult. This occurs even when substantial similarities exist in the process dynamics across different chemical systems. To address this challenge, we developed a novel time-series-transformer (TST) framework that exploits the powerful transfer learning capabilities inherent in transformer algorithms. This was demonstrated using readily available process data obtained from different crystallizers operating under various operational scenarios. Using this extensive dataset, we trained a TST model (CrystalGPT) to exhibit remarkable S2S transferability not only across all pre-established systems, but also to an unencountered system. CrystalGPT achieved a cumulative error across all systems, which is eight times superior to that of existing ML models. Additionally, we coupled CrystalGPT with a predictive controller to reduce the variance in setpoint tracking to just 1%.
Machine learning is typically framed from a perspective of i.i.d., and more importantly, isolated data. In parts, federated learning lifts this assumption, as it sets out to solve the real-world challenge of collaboratively learning a shared model from data distributed across clients. However, motivated primarily by privacy and computational constraints, the fact that data may change, distributions drift, or even tasks advance individually on clients, is seldom taken into account. The field of continual learning addresses this separate challenge and first steps have recently been taken to leverage synergies in distributed supervised settings, in which several clients learn to solve changing classification tasks over time without forgetting previously seen ones. Motivated by these prior works, we posit that such federated continual learning should be grounded in unsupervised learning of representations that are shared across clients; in the loose spirit of how humans can indirectly leverage others' experience without exposure to a specific task. For this purpose, we demonstrate that masked autoencoders for distribution estimation are particularly amenable to this setup. Specifically, their masking strategy can be seamlessly integrated with task attention mechanisms to enable selective knowledge transfer between clients. We empirically corroborate the latter statement through several continual federated scenarios on both image and binary datasets.