In this paper, we present an efficient solution for weed classification in agriculture. We focus on optimizing model performance at inference while respecting the constraints of the agricultural domain. We propose a Quantized Deep Neural Network model that classifies a dataset of 9 weed classes using 8-bit integer (int8) quantization, a departure from standard 32-bit floating point (fp32) models. Recognizing the hardware resource limitations in agriculture, our model balances model size, inference time, and accuracy, aligning with practical requirements. We evaluate the approach on ResNet-50 and InceptionV3 architectures, comparing their performance against their int8 quantized versions. Transfer learning and fine-tuning are applied using the DeepWeeds dataset. The results show staggering model size and inference time reductions while maintaining accuracy in real-world production scenarios like Desktop, Mobile and Raspberry Pi. Our work sheds light on a promising direction for efficient AI in agriculture, holding potential for broader applications. Code: https://github.com/parikshit14/QNN-for-weed
Pathological image analysis is a crucial field in computer-aided diagnosis, where deep learning is widely applied. Transfer learning using pre-trained models initialized on natural images has effectively improved the downstream pathological performance. However, the lack of sophisticated domain-specific pathological initialization hinders their potential. Self-supervised learning (SSL) enables pre-training without sample-level labels, which has great potential to overcome the challenge of expensive annotations. Thus, studies focusing on pathological SSL pre-training call for a comprehensive and standardized dataset, similar to the ImageNet in computer vision. This paper presents the comprehensive pathological image analysis (CPIA) dataset, a large-scale SSL pre-training dataset combining 103 open-source datasets with extensive standardization. The CPIA dataset contains 21,427,877 standardized images, covering over 48 organs/tissues and about 100 kinds of diseases, which includes two main data types: whole slide images (WSIs) and characteristic regions of interest (ROIs). A four-scale WSI standardization process is proposed based on the uniform resolution in microns per pixel (MPP), while the ROIs are divided into three scales artificially. This multi-scale dataset is built with the diagnosis habits under the supervision of experienced senior pathologists. The CPIA dataset facilitates a comprehensive pathological understanding and enables pattern discovery explorations. Additionally, to launch the CPIA dataset, several state-of-the-art (SOTA) baselines of SSL pre-training and downstream evaluation are specially conducted. The CPIA dataset along with baselines is available at https://github.com/zhanglab2021/CPIA_Dataset.
Conformal inference is a fundamental and versatile tool that provides distribution-free guarantees for many machine learning tasks. We consider the transductive setting, where decisions are made on a test sample of $m$ new points, giving rise to $m$ conformal $p$-values. While classical results only concern their marginal distribution, we show that their joint distribution follows a P\'olya urn model, and establish a concentration inequality for their empirical distribution function. The results hold for arbitrary exchangeable scores, including adaptive ones that can use the covariates of the test+calibration samples at training stage for increased accuracy. We demonstrate the usefulness of these theoretical results through uniform, in-probability guarantees for two machine learning tasks of current interest: interval prediction for transductive transfer learning and novelty detection based on two-class classification.
Synthetic data generation, a cornerstone of Generative Artificial Intelligence, promotes a paradigm shift in data science by addressing data scarcity and privacy while enabling unprecedented performance. As synthetic data becomes more prevalent, concerns emerge regarding the accuracy of statistical methods when applied to synthetic data in contrast to raw data. This article explores the effectiveness of statistical methods on synthetic data and the privacy risks of synthetic data. Regarding effectiveness, we present the Synthetic Data Generation for Analytics framework. This framework applies statistical approaches to high-quality synthetic data produced by generative models like tabular diffusion models, which, initially trained on raw data, benefit from insights from pertinent studies through transfer learning. A key finding within this framework is the generational effect, which reveals that the error rate of statistical methods on synthetic data decreases with the addition of more synthetic data but may eventually rise or stabilize. This phenomenon, stemming from the challenge of accurately mirroring raw data distributions, highlights a "reflection point"-an ideal volume of synthetic data defined by specific error metrics. Through three case studies, sentiment analysis, predictive modeling of structured data, and inference in tabular data, we validate the superior performance of this framework compared to conventional approaches. On privacy, synthetic data imposes lower risks while supporting the differential privacy standard. These studies underscore synthetic data's untapped potential in redefining data science's landscape.
Training deep networks requires various design decisions regarding for instance their architecture, data augmentation, or optimization. In this work, we find these training variations to result in networks learning unique feature sets from the data. Using public model libraries comprising thousands of models trained on canonical datasets like ImageNet, we observe that for arbitrary pairings of pretrained models, one model extracts significant data context unavailable in the other -- independent of overall performance. Given any arbitrary pairing of pretrained models and no external rankings (such as separate test sets, e.g. due to data privacy), we investigate if it is possible to transfer such "complementary" knowledge from one model to another without performance degradation -- a task made particularly difficult as additional knowledge can be contained in stronger, equiperformant or weaker models. Yet facilitating robust transfer in scenarios agnostic to pretrained model pairings would unlock auxiliary gains and knowledge fusion from any model repository without restrictions on model and problem specifics - including from weaker, lower-performance models. This work therefore provides an initial, in-depth exploration on the viability of such general-purpose knowledge transfer. Across large-scale experiments, we first reveal the shortcomings of standard knowledge distillation techniques, and then propose a much more general extension through data partitioning for successful transfer between nearly all pretrained models, which we show can also be done unsupervised. Finally, we assess both the scalability and impact of fundamental model properties on successful model-agnostic knowledge transfer.
Deploying deep learning on Synthetic Aperture Radar (SAR) data is becoming more common for mapping purposes. One such case is sea ice, which is highly dynamic and rapidly changes as a result of the combined effect of wind, temperature, and ocean currents. Therefore, frequent mapping of sea ice is necessary to ensure safe marine navigation. However, there is a general shortage of expert-labeled data to train deep learning algorithms. Fine-tuning a pre-trained model on SAR imagery is a potential solution. In this paper, we compare the performance of deep learning models trained from scratch using randomly initialized weights against pre-trained models that we fine-tune for this purpose. Our results show that pre-trained models lead to better results, especially on test samples from the melt season.
Large protein language models are adept at capturing the underlying evolutionary information in primary structures, offering significant practical value for protein engineering. Compared to natural language models, protein amino acid sequences have a smaller data volume and a limited combinatorial space. Choosing an appropriate vocabulary size to optimize the pre-trained model is a pivotal issue. Moreover, despite the wealth of benchmarks and studies in the natural language community, there remains a lack of a comprehensive benchmark for systematically evaluating protein language model quality. Given these challenges, PETA trained language models with 14 different vocabulary sizes under three tokenization methods. It conducted thousands of tests on 33 diverse downstream datasets to assess the models' transfer learning capabilities, incorporating two classification heads and three random seeds to mitigate potential biases. Extensive experiments indicate that vocabulary sizes between 50 and 200 optimize the model, whereas sizes exceeding 800 detrimentally affect the model's representational performance. Our code, model weights and datasets are available at https://github.com/ginnm/ProteinPretraining.
We assemble a dataset of Creative-Commons-licensed (CC) images, which we use to train a set of open diffusion models that are qualitatively competitive with Stable Diffusion 2 (SD2). This task presents two challenges: (1) high-resolution CC images lack the captions necessary to train text-to-image generative models; (2) CC images are relatively scarce. In turn, to address these challenges, we use an intuitive transfer learning technique to produce a set of high-quality synthetic captions paired with curated CC images. We then develop a data- and compute-efficient training recipe that requires as little as 3% of the LAION-2B data needed to train existing SD2 models, but obtains comparable quality. These results indicate that we have a sufficient number of CC images (~70 million) for training high-quality models. Our training recipe also implements a variety of optimizations that achieve ~3X training speed-ups, enabling rapid model iteration. We leverage this recipe to train several high-quality text-to-image models, which we dub the CommonCanvas family. Our largest model achieves comparable performance to SD2 on a human evaluation, despite being trained on our CC dataset that is significantly smaller than LAION and using synthetic captions for training. We release our models, data, and code at https://github.com/mosaicml/diffusion/blob/main/assets/common-canvas.md
Language models have been shown to be rich enough to encode fMRI activations of certain Regions of Interest in our Brains. Previous works have explored transfer learning from representations learned for popular natural language processing tasks for predicting brain responses. In our work, we improve the performance of such encoders by creating an ensemble model out of 10 popular Language Models (2 syntactic and 8 semantic). We beat the current baselines by 10% on average across all ROIs through our ensembling methods.
Short-term load forecasting (STLF) is crucial for the daily operation of power grids. However, the non-linearity, non-stationarity, and randomness characterizing electricity demand time series renders STLF a challenging task. Various forecasting approaches have been proposed for improving STLF, including neural network (NN) models which are trained using data from multiple electricity demand series that may not necessary include the target series. In the present study, we investigate the performance of this special case of STLF, called transfer learning (TL), by considering a set of 27 time series that represent the national day-ahead electricity demand of indicative European countries. We employ a popular and easy-to-implement NN model and perform a clustering analysis to identify similar patterns among the series and assist TL. In this context, two different TL approaches, with and without the clustering step, are compiled and compared against each other as well as a typical NN training setup. Our results demonstrate that TL can outperform the conventional approach, especially when clustering techniques are considered.
Dataset distillation plays a crucial role in creating compact datasets with similar training performance compared with original large-scale ones. This is essential for addressing the challenges of data storage and training costs. Prevalent methods facilitate knowledge transfer by matching the gradients, embedding distributions, or training trajectories of synthetic images with those of the sampled original images. Although there are various matching objectives, currently the strategy for selecting original images is limited to naive random sampling. We argue that random sampling overlooks the evenness of the selected sample distribution, which may result in noisy or biased matching targets. Besides, the sample diversity is also not constrained by random sampling. Additionally, current methods predominantly focus on single-dimensional matching, where information is not fully utilized. To address these challenges, we propose a novel matching strategy called Dataset Distillation by Bidirectional REpresentAtive Matching (DREAM+), which selects representative original images for bidirectional matching. DREAM+ is applicable to a variety of mainstream dataset distillation frameworks and significantly reduces the number of distillation iterations by more than 15 times without affecting performance. Given sufficient training time, DREAM+ can further improve the performance and achieve state-of-the-art results. We have released the code at github.com/NUS-HPC-AI-Lab/DREAM+.
There has been a significant surge of interest in visually grounded grammar induction in recent times. While a variety of models have been developed for the task and have demonstrated impressive performance, they have not been evaluated on text domains that are different from the training domain, so it is unclear if the improvements brought by visual groundings are transferable. Our study aims to fill this gap and assess the degree of transferability. We start by extending VC-PCFG (short for Visually-grounded Compound PCFG~\citep{zhao-titov-2020-visually}) in such a way that it can transfer across text domains. We consider a zero-shot transfer learning setting where a model is trained on the source domain and is directly applied to target domains, without any further training. Our experimental results suggest that: the benefits from using visual groundings transfer to text in a domain similar to the training domain but fail to transfer to remote domains. Further, we conduct data and result analysis; we find that the lexicon overlap between the source domain and the target domain is the most important factor in the transferability of VC-PCFG.
Few-shot learning or meta-learning leverages the data scarcity problem in machine learning. Traditionally, training data requires a multitude of samples and labeling for supervised learning. To address this issue, we propose a one-shot unsupervised meta-learning to learn the latent representation of the training samples. We use augmented samples as the query set during the training phase of the unsupervised meta-learning. A temperature-scaled cross-entropy loss is used in the inner loop of meta-learning to prevent overfitting during unsupervised learning. The learned parameters from this step are applied to the targeted supervised meta-learning in a transfer-learning fashion for initialization and fast adaptation with improved accuracy. The proposed method is model agnostic and can aid any meta-learning model to improve accuracy. We use model agnostic meta-learning (MAML) and relation network (RN) on Omniglot and mini-Imagenet datasets to demonstrate the performance of the proposed method. Furthermore, a meta-learning model with the proposed initialization can achieve satisfactory accuracy with significantly fewer training samples.
Continual pre-training has been urgent for adapting a pre-trained model to a multitude of domains and tasks in the fast-evolving world. In practice, a continually pre-trained model is expected to demonstrate not only greater capacity when fine-tuned on pre-trained domains but also a non-decreasing performance on unseen ones. In this work, we first investigate such anytime fine-tuning effectiveness of existing continual pre-training approaches, concluding with unanimously decreased performance on unseen domains. To this end, we propose a prompt-guided continual pre-training method, where we train a hypernetwork to generate domain-specific prompts by both agreement and disagreement losses. The agreement loss maximally preserves the generalization of a pre-trained model to new domains, and the disagreement one guards the exclusiveness of the generated hidden states for each domain. Remarkably, prompts by the hypernetwork alleviate the domain identity when fine-tuning and promote knowledge transfer across domains. Our method achieved improvements of 3.57% and 3.4% on two real-world datasets (including domain shift and temporal shift), respectively, demonstrating its efficacy.
Large language models (LLMs) have exhibited considerable cross-lingual generalization abilities, whereby they implicitly transfer knowledge across languages. However, the transfer is not equally successful for all languages, especially for low-resource ones, which poses an ongoing challenge. It is unclear whether we have reached the limits of implicit cross-lingual generalization and if explicit knowledge transfer is viable. In this paper, we investigate the potential for explicitly aligning conceptual correspondence between languages to enhance cross-lingual generalization. Using the syntactic aspect of language as a testbed, our analyses of 43 languages reveal a high degree of alignability among the spaces of structural concepts within each language for both encoder-only and decoder-only LLMs. We then propose a meta-learning-based method to learn to align conceptual spaces of different languages, which facilitates zero-shot and few-shot generalization in concept classification and also offers insights into the cross-lingual in-context learning phenomenon. Experiments on syntactic analysis tasks show that our approach achieves competitive results with state-of-the-art methods and narrows the performance gap between languages, particularly benefiting those with limited resources.
Machine learning strategies like multi-task learning, meta-learning, and transfer learning enable efficient adaptation of machine learning models to specific applications in healthcare, such as prediction of various diseases, by leveraging generalizable knowledge across large datasets and multiple domains. In particular, Bayesian meta-learning methods pool data across related prediction tasks to learn prior distributions for model parameters, which are then used to derive models for specific tasks. However, inter- and intra-task variability due to disease heterogeneity and other patient-level differences pose challenges of negative transfer during shared learning and poor generalizability to new patients. We introduce a novel Bayesian meta-learning approach that aims to address this in two key settings: (1) predictions for new patients (same population as the training set) and (2) adapting to new patient populations. Our main contribution is in modeling similarity between causal mechanisms of the tasks, for (1) mitigating negative transfer during training and (2) fine-tuning that pools information from tasks that are expected to aid generalizability. We propose an algorithm for implementing this approach for Bayesian deep learning, and apply it to a case study for stroke prediction tasks using electronic health record data. Experiments for the UK Biobank dataset as the training population demonstrated significant generalizability improvements compared to standard meta-learning, non-causal task similarity measures, and local baselines (separate models for each task). This was assessed for a variety of tasks that considered both new patients from the training population (UK Biobank) and a new population (FinnGen).
High-resolution 3D object generation remains a challenging task primarily due to the limited availability of comprehensive annotated training data. Recent advancements have aimed to overcome this constraint by harnessing image generative models, pretrained on extensive curated web datasets, using knowledge transfer techniques like Score Distillation Sampling (SDS). Efficiently addressing the requirements of high-resolution rendering often necessitates the adoption of latent representation-based models, such as the Latent Diffusion Model (LDM). In this framework, a significant challenge arises: To compute gradients for individual image pixels, it is necessary to backpropagate gradients from the designated latent space through the frozen components of the image model, such as the VAE encoder used within LDM. However, this gradient propagation pathway has never been optimized, remaining uncontrolled during training. We find that the unregulated gradients adversely affect the 3D model's capacity in acquiring texture-related information from the image generative model, leading to poor quality appearance synthesis. To address this overarching challenge, we propose an innovative operation termed Pixel-wise Gradient Clipping (PGC) designed for seamless integration into existing 3D generative models, thereby enhancing their synthesis quality. Specifically, we control the magnitude of stochastic gradients by clipping the pixel-wise gradients efficiently, while preserving crucial texture-related gradient directions. Despite this simplicity and minimal extra cost, extensive experiments demonstrate the efficacy of our PGC in enhancing the performance of existing 3D generative models for high-resolution object rendering.
Localization in GPS-denied outdoor locations, such as street canyons in an urban or metropolitan environment, has many applications. Machine Learning (ML) is widely used to tackle this critical problem. One challenge lies in the mixture of line-of-sight (LOS), obstructed LOS (OLOS), and non-LOS (NLOS) conditions. In this paper, we consider a semantic localization that treats these three propagation conditions as the ''semantic objects", and aims to determine them together with the actual localization, and show that this increases accuracy and robustness. Furthermore, the propagation conditions are highly dynamic, since obstruction by cars or trucks can change the channel state information (CSI) at a fixed location over time. We therefore consider the blockage by such dynamic objects as another semantic state. Based on these considerations, we formulate the semantic localization with a joint task (coordinates regression and semantics classification) learning problem. Another problem created by the dynamics is the fact that each location may be characterized by a number of different CSIs. To avoid the need for excessive amount of labeled training data, we propose a multi-task deep domain adaptation (DA) based localization technique, training neural networks with a limited number of labeled samples and numerous unlabeled ones. Besides, we introduce novel scenario adaptive learning strategies to ensure efficient representation learning and successful knowledge transfer. Finally, we use Bayesian theory for uncertainty modeling of the importance weights in each task, reducing the need for time-consuming parameter finetuning; furthermore, with some mild assumptions, we derive the related log-likelihood for the joint task and present the deep homoscedastic DA based localization method.
Parameter-efficient fine-tuning (PEFT) has shown its effectiveness in adapting the pre-trained language models to downstream tasks while only updating a small number of parameters. Despite the success, most existing methods independently adapt to each task without considering knowledge transfer between tasks and are limited to low-data regimes. To overcome this issue, we propose Prototype-based HyperAdapter (PHA), a novel framework built on the adapter-tuning and hypernetwork. It introduces an instance-dense retriever and a prototypical hypernetwork to generate the conditional modules in a sample-efficient manner. This leads to comparable performance improvements against existing PEFT methods on multi-task learning and few-shot transfer learning. More importantly, when the available data size gets smaller, our method outperforms other strong baselines by a large margin. Based on our extensive empirical experiments across various datasets, we demonstrate that PHA strikes a better trade-off between trainable parameters, accuracy on stream tasks, and sample efficiency.
Large Language Models (LLMs) inherently encode a wealth of knowledge within their parameters through pre-training on extensive corpora. While prior research has delved into operations on these parameters to manipulate the underlying implicit knowledge (encompassing detection, editing, and merging), there remains an ambiguous understanding regarding their transferability across models with varying scales. In this paper, we seek to empirically investigate knowledge transfer from larger to smaller models through a parametric perspective. To achieve this, we employ sensitivity-based techniques to extract and align knowledge-specific parameters between different LLMs. Moreover, the LoRA module is used as the intermediary mechanism for injecting the extracted knowledge into smaller models. Evaluations across four benchmarks validate the efficacy of our proposed method. Our findings highlight the critical factors contributing to the process of parametric knowledge transfer, underscoring the transferability of model parameters across LLMs of different scales. Project website: https://maszhongming.github.io/ParaKnowTransfer.