Learning fine-scale details of a coastal ocean simulation from a coarse representation is a challenging task. For real-world applications, high-resolution simulations are necessary to advance understanding of many coastal processes, specifically, to predict flooding resulting from tsunamis and storm surges. We propose a Deep Network for Coastal Super-Resolution (DNCSR) for spatiotemporal enhancement to efficiently learn the high-resolution numerical solution. Given images of coastal simulations produced on low-resolution computational meshes using low polynomial order discontinuous Galerkin discretizations and a coarse temporal resolution, the proposed DNCSR learns to produce high-resolution free surface elevation and velocity visualizations in both time and space. To efficiently model the dynamic changes over time and space, we propose grid-aware spatiotemporal attention to project the temporal features to the spatial domain for non-local feature matching. The coordinate information is also utilized via positional encoding. For the final reconstruction, we use the spatiotemporal bilinear operation to interpolate the missing frames and then expand the feature maps to the frequency domain for residual mapping. Besides data-driven losses, the proposed physics-informed loss guarantees gradient consistency and momentum changes. Their combination contributes to the overall 24% improvements in RMSE. To train the proposed model, we propose a large-scale coastal simulation dataset and use it for model optimization and evaluation. Our method shows superior super-resolution quality and fast computation compared to the state-of-the-art methods.
Variational continual learning (VCL) is a turn-key learning algorithm that has state-of-the-art performance among the best continual learning models. In our work, we explore an extension of the generalized variational continual learning (GVCL) model, named AutoVCL, which combines task heuristics for informed learning and model optimization. We demonstrate that our model outperforms the standard GVCL with fixed hyperparameters, benefiting from the automatic adjustment of the hyperparameter based on the difficulty and similarity of the incoming task compared to the previous tasks.
Reinforcement learning is used to align language models with human preference signals after first pre-training the model to predict the next token of text within a large corpus using likelihood maximization. Before being deployed in a specific domain, models are often further fine-tuned on task specific data. Since human preferences are often unavailable for the last step, it is performed using likelihood maximization as that is the typical default method. However, reinforcement learning has other advantages besides facilitating alignment to a human derived reward function. For one, whereas likelihood maximization is a form of imitation learning in which the model is trained on what to do under ideal conditions, reinforcement learning is not limited to demonstrating actions just for optimally reached states and trains a model what to do under a range of scenarios as it explores the policy space. In addition, it also trains a model what not to do, suppressing competitive but poor actions. This work develops a framework for last-mile fine-tuning using reinforcement learning and tests whether it garners performance gains. The experiments center on abstractive summarization, but the framework is general and broadly applicable. Use of the procedure produced significantly better results than likelihood maximization when comparing raw predictions. For the specific data tested, the gap could be bridged by employing post-processing of the maximum likelihood outputs. Nonetheless, the framework offers a new avenue for model optimization in situations where post-processing may be less straightforward or effective, and it can be extended to include more complex classes of undesirable outputs to penalize and train against, such as hallucinations.
Machine learning models are often criticized for their black-box nature, raising concerns about their applicability in critical decision-making scenarios. Consequently, there is a growing demand for interpretable models in such contexts. In this study, we introduce Model-based Deep Rule Forests (mobDRF), an interpretable representation learning algorithm designed to extract transparent models from data. By leveraging IF-THEN rules with multi-level logic expressions, mobDRF enhances the interpretability of existing models without compromising accuracy. We apply mobDRF to identify key risk factors for cognitive decline in an elderly population, demonstrating its effectiveness in subgroup analysis and local model optimization. Our method offers a promising solution for developing trustworthy and interpretable machine learning models, particularly valuable in fields like healthcare, where understanding differential effects across patient subgroups can lead to more personalized and effective treatments.
Smaller machine learning models, with less complex architectures and sensor inputs, can benefit wearable sensor-based human activity recognition (HAR) systems in many ways, from complexity and cost to battery life. In the specific case of smart factories, optimizing human-robot collaboration hinges on the implementation of cutting-edge, human-centric AI systems. To this end, workers' activity recognition enables accurate quantification of performance metrics, improving efficiency holistically. We present a two-stage semantic-aware knowledge distillation (KD) approach, TSAK, for efficient, privacy-aware, and wearable HAR in manufacturing lines, which reduces the input sensor modalities as well as the machine learning model size, while reaching similar recognition performance as a larger multi-modal and multi-positional teacher model. The first stage incorporates a teacher classifier model encoding attention, causal, and combined representations. The second stage encompasses a semantic classifier merging the three representations from the first stage. To evaluate TSAK, we recorded a multi-modal dataset at a smart factory testbed with wearable and privacy-aware sensors (IMU and capacitive) located on both workers' hands. In addition, we evaluated our approach on OpenPack, the only available open dataset mimicking the wearable sensor placements on both hands in the manufacturing HAR scenario. We compared several KD strategies with different representations to regulate the training process of a smaller student model. Compared to the larger teacher model, the student model takes fewer sensor channels from a single hand, has 79% fewer parameters, runs 8.88 times faster, and requires 96.6% less computing power (FLOPS).
The current mainstream multi-modal medical image-to-image translation methods face a contradiction. Supervised methods with outstanding performance rely on pixel-wise aligned training data to constrain the model optimization. However, obtaining pixel-wise aligned multi-modal medical image datasets is challenging. Unsupervised methods can be trained without paired data, but their reliability cannot be guaranteed. At present, there is no ideal multi-modal medical image-to-image translation method that can generate reliable translation results without the need for pixel-wise aligned data. This work aims to develop a novel medical image-to-image translation model that is independent of pixel-wise aligned data (MITIA), enabling reliable multi-modal medical image-to-image translation under the condition of misaligned training data. The proposed MITIA model utilizes a prior extraction network composed of a multi-modal medical image registration module and a multi-modal misalignment error detection module to extract pixel-level prior information from training data with misalignment errors to the largest extent. The extracted prior information is then used to construct a regularization term to constrain the optimization of the unsupervised cycle-consistent GAN model, restricting its solution space and thereby improving the performance and reliability of the generator. We trained the MITIA model using six datasets containing different misalignment errors and two well-aligned datasets. Subsequently, we compared the proposed method with six other state-of-the-art image-to-image translation methods. The results of both quantitative analysis and qualitative visual inspection indicate that MITIA achieves superior performance compared to the competing state-of-the-art methods, both on misaligned data and aligned data.
In this paper, we introduce Nested Low-Rank Adaptation (NoRA), a novel approach to parameter-efficient fine-tuning that extends the capabilities of Low-Rank Adaptation (LoRA) techniques. Vanilla LoRA overlooks pre-trained weight inheritance and still requires fine-tuning numerous parameters. To addresses these issues, our NoRA adopts a dual-layer nested structure with Singular Value Decomposition (SVD), effectively leveraging original matrix knowledge while reducing tunable parameters. Specifically, NoRA freezes the outer LoRA weights and utilizes an inner LoRA design, providing enhanced control over model optimization. This approach allows the model to more precisely adapt to specific tasks while maintaining a compact parameter space. By freezing outer LoRA weights and using an inner LoRA design, NoRA enables precise task adaptation with a compact parameter space. Evaluations on tasks including commonsense reasoning with large language models, fine-tuning vision-language models, and subject-driven generation demonstrate NoRA's superiority over LoRA and its variants. Code will be released upon acceptance.
Dose prediction plays a key role in knowledge-based planning (KBP) by automatically generating patient-specific dose distribution. Recent advances in deep learning-based dose prediction methods necessitates collaboration among data contributors for improved performance. Federated learning (FL) has emerged as a solution, enabling medical centers to jointly train deep-learning models without compromising patient data privacy. We developed the FedKBP framework to evaluate the performances of centralized, federated, and individual (i.e. separated) training of dose prediction model on the 340 plans from OpenKBP dataset. To simulate FL and individual training, we divided the data into 8 training sites. To evaluate the effect of inter-site data variation on model training, we implemented two types of case distributions: 1) Independent and identically distributed (IID), where the training and validating cases were evenly divided among the 8 sites, and 2) non-IID, where some sites have more cases than others. The results show FL consistently outperforms individual training on both model optimization speed and out-of-sample testing scores, highlighting the advantage of FL over individual training. Under IID data division, FL shows comparable performance to centralized training, underscoring FL as a promising alternative to traditional pooled-data training. Under non-IID division, larger sites outperformed smaller sites by up to 19% on testing scores, confirming the need of collaboration among data owners to achieve better prediction accuracy. Meanwhile, non-IID FL showed reduced performance as compared to IID FL, posing the need for more sophisticated FL method beyond mere model averaging to handle data variation among participating sites.
Audio-driven 3D face animation is increasingly vital in live streaming and augmented reality applications. While remarkable progress has been observed, most existing approaches are designed for specific individuals with predefined speaking styles, thus neglecting the adaptability to varied speaking styles. To address this limitation, this paper introduces MetaFace, a novel methodology meticulously crafted for speaking style adaptation. Grounded in the novel concept of meta-learning, MetaFace is composed of several key components: the Robust Meta Initialization Stage (RMIS) for fundamental speaking style adaptation, the Dynamic Relation Mining Neural Process (DRMN) for forging connections between observed and unobserved speaking styles, and the Low-rank Matrix Memory Reduction Approach to enhance the efficiency of model optimization as well as learning style details. Leveraging these novel designs, MetaFace not only significantly outperforms robust existing baselines but also establishes a new state-of-the-art, as substantiated by our experimental results.
Optimizing Cox regression and its neural network variants poses substantial computational challenges in large-scale studies. Stochastic gradient descent (SGD), known for its scalability in model optimization, has recently been adapted to optimize Cox models. Unlike its conventional application, which typically targets a sum of independent individual loss, SGD for Cox models updates parameters based on the partial likelihood of a subset of data. Despite its empirical success, the theoretical foundation for optimizing Cox partial likelihood with SGD is largely underexplored. In this work, we demonstrate that the SGD estimator targets an objective function that is batch-size-dependent. We establish that the SGD estimator for the Cox neural network (Cox-NN) is consistent and achieves the optimal minimax convergence rate up to a polylogarithmic factor. For Cox regression, we further prove the $\sqrt{n}$-consistency and asymptotic normality of the SGD estimator, with variance depending on the batch size. Furthermore, we quantify the impact of batch size on Cox-NN training and its effect on the SGD estimator's asymptotic efficiency in Cox regression. These findings are validated by extensive numerical experiments and provide guidance for selecting batch sizes in SGD applications. Finally, we demonstrate the effectiveness of SGD in a real-world application where GD is unfeasible due to the large scale of data.
Painting classification plays a vital role in organizing, finding, and suggesting artwork for digital and classic art galleries. Existing methods struggle with adapting knowledge from the real world to artistic images during training, leading to poor performance when dealing with different datasets. Our innovation lies in addressing these challenges through a two-step process. First, we generate more data using Style Transfer with Adaptive Instance Normalization (AdaIN), bridging the gap between diverse styles. Then, our classifier gains a boost with feature-map adaptive spatial attention modules, improving its understanding of artistic details. Moreover, we tackle the problem of imbalanced class representation by dynamically adjusting augmented samples. Through a dual-stage process involving careful hyperparameter search and model fine-tuning, we achieve an impressive 87.24\% accuracy using the ResNet-50 backbone over 40 training epochs. Our study explores quantitative analyses that compare different pretrained backbones, investigates model optimization through ablation studies, and examines how varying augmentation levels affect model performance. Complementing this, our qualitative experiments offer valuable insights into the model's decision-making process using spatial attention and its ability to differentiate between easy and challenging samples based on confidence ranking.
Investigating the genetic architecture of complex diseases is challenging due to the multifactorial and interactive landscape of genomic and environmental influences. Although genome-wide association studies (GWAS) have identified thousands of variants for multiple complex traits, conventional statistical approaches can be limited by simplified assumptions such as linearity and lack of epistasis in models. In this work, we trained artificial neural networks to predict complex traits using both simulated and real genotype-phenotype datasets. We extracted feature importance scores via different post hoc interpretability methods to identify potentially associated loci (PAL) for the target phenotype and devised an approach for obtaining p-values for the detected PAL. Simulations with various parameters demonstrated that associated loci can be detected with good precision using strict selection criteria. By applying our approach to the schizophrenia cohort in the Estonian Biobank, we detected multiple loci associated with this highly polygenic and heritable disorder. There was significant concordance between PAL and loci previously associated with schizophrenia and bipolar disorder, with enrichment analyses of genes within the identified PAL predominantly highlighting terms related to brain morphology and function. With advancements in model optimization and uncertainty quantification, artificial neural networks have the potential to enhance the identification of genomic loci associated with complex diseases, offering a more comprehensive approach for GWAS and serving as initial screening tools for subsequent functional studies.
This study introduces novel superior scoring rules called Penalized Brier Score (PBS) and Penalized Logarithmic Loss (PLL) to improve model evaluation for probabilistic classification. Traditional scoring rules like Brier Score and Logarithmic Loss sometimes assign better scores to misclassifications in comparison with correct classifications. This discrepancy from the actual preference for rewarding correct classifications can lead to suboptimal model selection. By integrating penalties for misclassifications, PBS and PLL modify traditional proper scoring rules to consistently assign better scores to correct predictions. Formal proofs demonstrate that PBS and PLL satisfy strictly proper scoring rule properties while also preferentially rewarding accurate classifications. Experiments showcase the benefits of using PBS and PLL for model selection, model checkpointing, and early stopping. PBS exhibits a higher negative correlation with the F1 score compared to the Brier Score during training. Thus, PBS more effectively identifies optimal checkpoints and early stopping points, leading to improved F1 scores. Comparative analysis verifies models selected by PBS and PLL achieve superior F1 scores. Therefore, PBS and PLL address the gap between uncertainty quantification and accuracy maximization by encapsulating both proper scoring principles and explicit preference for true classifications. The proposed metrics can enhance model evaluation and selection for reliable probabilistic classification.
Despite the substantial progress in deep learning, its adoption in industrial robotics projects remains limited, primarily due to challenges in data acquisition and labeling. Previous sim2real approaches using domain randomization require extensive scene and model optimization. To address these issues, we introduce an innovative physically-based structured light simulation system, generating both RGB and physically realistic depth images, surpassing previous dataset generation tools. We create an RGBD dataset tailored for robotic industrial grasping scenarios and evaluate it across various tasks, including object detection, instance segmentation, and embedding sim2real visual perception in industrial robotic grasping. By reducing the sim2real gap and enhancing deep learning training, we facilitate the application of deep learning models in industrial settings. Project details are available at https://baikaixinpublic.github.io/structured light 3D synthesizer/.
Segmenting tumors in histological images is vital for cancer diagnosis. While fully supervised models excel with pixel-level annotations, creating such annotations is labor-intensive and costly. Accurate histopathology image segmentation under weakly-supervised conditions with coarse-grained image labels is still a challenging problem. Although multiple instance learning (MIL) has shown promise in segmentation tasks, surprisingly, no previous pseudo-supervision methods have used MIL-based outputs as pseudo-masks for training. We suspect this stems from concerns over noises in MIL results affecting pseudo supervision quality. To explore the potential of leveraging MIL-based segmentation for pseudo supervision, we propose a novel distillation framework for histopathology image segmentation. This framework introduces a iterative fusion-knowledge distillation strategy, enabling the student model to learn directly from the teacher's comprehensive outcomes. Through dynamic role reversal between the fixed teacher and learnable student models and the incorporation of weighted cross-entropy loss for model optimization, our approach prevents performance deterioration and noise amplification during knowledge distillation. Experimental results on public histopathology datasets, Camelyon16 and Digestpath2019, demonstrate that our approach not only complements various MIL-based segmentation methods but also significantly enhances their performance. Additionally, our method achieves new SOTA in the field.
Training modern LLMs is extremely resource intensive, and customizing them for various deployment scenarios characterized by limited compute and memory resources through repeated training is impractical. In this paper, we introduce Flextron, a network architecture and post-training model optimization framework supporting flexible model deployment. The Flextron architecture utilizes a nested elastic structure to rapidly adapt to specific user-defined latency and accuracy targets during inference with no additional fine-tuning required. It is also input-adaptive, and can automatically route tokens through its sub-networks for improved performance and efficiency. We present a sample-efficient training method and associated routing algorithms for systematically transforming an existing trained LLM into a Flextron model. We evaluate Flextron on the GPT-3 and LLama-2 family of LLMs, and demonstrate superior performance over multiple end-to-end trained variants and other state-of-the-art elastic networks, all with a single pretraining run that consumes a mere 7.63% tokens compared to original pretraining.
In recent years, many incomplete multi-view clustering methods have been proposed to address the challenging and new clustering task on incomplete multi-view data whose part of view representations are not fully collected for some samples. Although extensive experiments have validated the effectiveness of these methods for handling the incomplete learning issue, a common issue exists, i.e., these methods all ignore the discriminative/important difference of discriminative features and noisy features. In this paper, to address the above issue, a new incomplete multi-view clustering model, called Graph Regularized and fEature Aware maTrix Factorization (GreatF), is proposed. Different from the existing methods, we introduce an adaptive feature weighting constraint to the matrix factorization-based multi-view representation learning model. With this weighting constraint, the effect of the discriminative features can be enhanced while the negative effect caused by the redundant and noisy features can be eliminated for the model optimization; thus, the robustness of the model can be enhanced. In addition, in this work, we designed a new graph-embedded consensus representation learning term in which consensus representation learning and structure information preservation are integrated into a joint model with one term. In particular, this term provides a more concise approach to obtain the structured consensus representation from incomplete multi-view data. Experimental results on four well-known datasets demonstrate that GreatF performs better than the state-of-the-art incomplete multi-view clustering methods.
In this research, we introduce an innovative method for synthesizing medical images using generative adversarial networks (GANs). Our proposed GANs method demonstrates the capability to produce realistic synthetic images even when trained on a limited quantity of real medical image data, showcasing commendable generalization prowess. To achieve this, we devised a generator and discriminator network architecture founded on deep convolutional neural networks (CNNs), leveraging the adversarial training paradigm for model optimization. Through extensive experimentation across diverse medical image datasets, our method exhibits robust performance, consistently generating synthetic images that closely emulate the structural and textural attributes of authentic medical images.
Recent research has shown that pruning large-scale language models for inference is an effective approach to improving model efficiency, significantly reducing model weights with minimal impact on performance. Interestingly, pruning can sometimes even enhance accuracy by removing noise that accumulates during training, particularly through matrix decompositions. However, recent work has primarily focused on single matrix decompositions or lower precision techniques, which may fail to fully capture structural patterns. To address these limitations, we introduce TRAWL (Tensor Reduced and Approximated Weights for Large Language Models), a technique that applies tensor decomposition across multiple weight matrices to effectively denoise LLMs by capturing global structural patterns. Our experiments show that TRAWL improves model performance by up to 16% over baseline models on benchmark datasets, without requiring additional data, training, or fine-tuning.
The proliferation of complex deep learning (DL) models has revolutionized various applications, including computer vision-based solutions, prompting their integration into real-time systems. However, the resource-intensive nature of these models poses challenges for deployment on low-computational power and low-memory devices, like embedded and edge devices. This work empirically investigates the optimization of such complex DL models to analyze their functionality on an embedded device, particularly on the NVIDIA Jetson Nano. It evaluates the effectiveness of the optimized models in terms of their inference speed for image classification and video action detection. The experimental results reveal that, on average, optimized models exhibit a 16.11% speed improvement over their non-optimized counterparts. This not only emphasizes the critical need to consider hardware constraints and environmental sustainability in model development and deployment but also underscores the pivotal role of model optimization in enabling the widespread deployment of AI-assisted technologies on resource-constrained computational systems. It also serves as proof that prioritizing hardware-specific model optimization leads to efficient and scalable solutions that substantially decrease energy consumption and carbon footprint.