Model Selection with Model Zoo via Graph Learning

Pre-trained deep learning (DL) models are increasingly accessible in public repositories, i.e., model zoos. Given a new prediction task, finding the best model to fine-tune can be computationally intensive and costly, especially when the number of pre-trained models is large. Selecting the right pre-trained models is crucial, yet complicated by the diversity of models from various model families (like ResNet, Vit, Swin) and the hidden relationships between models and datasets. Existing methods, which utilize basic information from models and datasets to compute scores indicating model performance on target datasets, overlook the intrinsic relationships, limiting their effectiveness in model selection. In this study, we introduce TransferGraph, a novel framework that reformulates model selection as a graph learning problem. TransferGraph constructs a graph using extensive metadata extracted from models and datasets, while capturing their inherent relationships. Through comprehensive experiments across 16 real datasets, both images and texts, we demonstrate TransferGraph's effectiveness in capturing essential model-dataset relationships, yielding up to a 32% improvement in correlation between predicted performance and the actual fine-tuning results compared to the state-of-the-art methods.

Mutation-based Consistency Testing for Evaluating the Code Understanding Capability of LLMs

Large Language Models (LLMs) have shown remarkable capabilities in processing both natural and programming languages, which have enabled various applications in software engineering, such as requirement engineering, code generation, and software testing. However, existing code generation benchmarks do not necessarily assess the code understanding performance of LLMs, especially for the subtle inconsistencies that may arise between code and its semantics described in natural language. In this paper, we propose a novel method to systematically assess the code understanding performance of LLMs, particularly focusing on subtle differences between code and its descriptions, by introducing code mutations to existing code generation datasets. Code mutations are small changes that alter the semantics of the original code, creating a mismatch with the natural language description. We apply different types of code mutations, such as operator replacement and statement deletion, to generate inconsistent code-description pairs. We then use these pairs to test the ability of LLMs to correctly detect the inconsistencies. We propose a new LLM testing method, called Mutation-based Consistency Testing (MCT), and conduct a case study on the two popular LLMs, GPT-3.5 and GPT-4, using the state-of-the-art code generation benchmark, HumanEval-X, which consists of six programming languages (Python, C++, Java, Go, JavaScript, and Rust). We compare the performance of the LLMs across different types of code mutations and programming languages and analyze the results. We find that the LLMs show significant variation in their code understanding performance and that they have different strengths and weaknesses depending on the mutation type and language.

Self-navigated 3D diffusion MRI using an optimized CAIPI sampling and structured low-rank reconstruction

3D multi-slab acquisitions are an appealing approach for diffusion MRI because they are compatible with the imaging regime delivering optimal SNR efficiency. In conventional 3D multi-slab imaging, shot-to-shot phase variations caused by motion pose challenges due to the use of multi-shot k-space acquisition. Navigator acquisition after each imaging echo is typically employed to correct phase variations, which prolongs scan time and increases the specific absorption rate (SAR). The aim of this study is to develop a highly efficient, self-navigated method to correct for phase variations in 3D multi-slab diffusion MRI without explicitly acquiring navigators. The sampling of each shot is carefully designed to intersect with the central kz plane of each slab, and the multi-shot sampling is optimized for self-navigation performance while retaining decent reconstruction quality. The central kz intersections from all shots are jointly used to reconstruct a 2D phase map for each shot using a structured low-rank constrained reconstruction that leverages the redundancy in shot and coil dimensions. The phase maps are used to eliminate the shot-to-shot phase inconsistency in the final 3D multi-shot reconstruction. We demonstrate the method's efficacy using retrospective simulations and prospectively acquired in-vivo experiments at 1.22 mm and 1.09 mm isotropic resolutions. Compared to conventional navigated 3D multi-slab imaging, the proposed self-navigated method achieves comparable image quality while shortening the scan time by 31.7% and improving the SNR efficiency by 15.5%. The proposed method produces comparable quality of DTI and white matter tractography to conventional navigated 3D multi-slab acquisition with a much shorter scan time.

GPA-3D: Geometry-aware Prototype Alignment for Unsupervised Domain Adaptive 3D Object Detection from Point Clouds

LiDAR-based 3D detection has made great progress in recent years. However, the performance of 3D detectors is considerably limited when deployed in unseen environments, owing to the severe domain gap problem. Existing domain adaptive 3D detection methods do not adequately consider the problem of the distributional discrepancy in feature space, thereby hindering generalization of detectors across domains. In this work, we propose a novel unsupervised domain adaptive \textbf{3D} detection framework, namely \textbf{G}eometry-aware \textbf{P}rototype \textbf{A}lignment (\textbf{GPA-3D}), which explicitly leverages the intrinsic geometric relationship from point cloud objects to reduce the feature discrepancy, thus facilitating cross-domain transferring. Specifically, GPA-3D assigns a series of tailored and learnable prototypes to point cloud objects with distinct geometric structures. Each prototype aligns BEV (bird's-eye-view) features derived from corresponding point cloud objects on source and target domains, reducing the distributional discrepancy and achieving better adaptation. The evaluation results obtained on various benchmarks, including Waymo, nuScenes and KITTI, demonstrate the superiority of our GPA-3D over the state-of-the-art approaches for different adaptation scenarios. The MindSpore version code will be publicly available at \url{https://github.com/Liz66666/GPA3D}.

Metadata Representations for Queryable ML Model Zoos

Machine learning (ML) practitioners and organizations are building model zoos of pre-trained models, containing metadata describing properties of the ML models and datasets that are useful for reporting, auditing, reproducibility, and interpretability purposes. The metatada is currently not standardised; its expressivity is limited; and there is no interoperable way to store and query it. Consequently, model search, reuse, comparison, and composition are hindered. In this paper, we advocate for standardized ML model meta-data representation and management, proposing a toolkit supported to help practitioners manage and query that metadata.

Quantifying the uncertainty of neural networks using Monte Carlo dropout for deep learning based quantitative MRI

Dropout is conventionally used during the training phase as regularization method and for quantifying uncertainty in deep learning. We propose to use dropout during training as well as inference steps, and average multiple predictions to improve the accuracy, while reducing and quantifying the uncertainty. The results are evaluated for fractional anisotropy (FA) and mean diffusivity (MD) maps which are obtained from only 3 direction scans. With our method, accuracy can be improved significantly compared to network outputs without dropout, especially when the training dataset is small. Moreover, confidence maps are generated which may aid in diagnosis of unseen pathology or artifacts.

SDnDTI: Self-supervised deep learning-based denoising for diffusion tensor MRI

The noise in diffusion-weighted images (DWIs) decreases the accuracy and precision of diffusion tensor magnetic resonance imaging (DTI) derived microstructural parameters and leads to prolonged acquisition time for achieving improved signal-to-noise ratio (SNR). Deep learning-based image denoising using convolutional neural networks (CNNs) has superior performance but often requires additional high-SNR data for supervising the training of CNNs, which reduces the practical feasibility. We develop a self-supervised deep learning-based method entitled "SDnDTI" for denoising DTI data, which does not require additional high-SNR data for training. Specifically, SDnDTI divides multi-directional DTI data into many subsets, each consisting of six DWI volumes along optimally chosen diffusion-encoding directions that are robust to noise for the tensor fitting, and then synthesizes DWI volumes along all acquired directions from the diffusion tensors fitted using each subset of the data as the input data of CNNs. On the other hand, SDnDTI synthesizes DWI volumes along acquired diffusion-encoding directions with higher SNR from the diffusion tensors fitted using all acquired data as the training target. SDnDTI removes noise from each subset of synthesized DWI volumes using a deep 3-dimensional CNN to match the quality of the cleaner target DWI volumes and achieves even higher SNR by averaging all subsets of denoised data. The denoising efficacy of SDnDTI is demonstrated on two datasets provided by the Human Connectome Project (HCP) and the Lifespan HCP in Aging. The SDnDTI results preserve image sharpness and textural details and substantially improve upon those from the raw data. The results of SDnDTI are comparable to those from supervised learning-based denoising and outperform those from state-of-the-art conventional denoising algorithms including BM4D, AONLM and MPPCA.

Towards Unified Surgical Skill Assessment

Surgical skills have a great influence on surgical safety and patients' well-being. Traditional assessment of surgical skills involves strenuous manual efforts, which lacks efficiency and repeatability. Therefore, we attempt to automatically predict how well the surgery is performed using the surgical video. In this paper, a unified multi-path framework for automatic surgical skill assessment is proposed, which takes care of multiple composing aspects of surgical skills, including surgical tool usage, intraoperative event pattern, and other skill proxies. The dependency relationships among these different aspects are specially modeled by a path dependency module in the framework. We conduct extensive experiments on the JIGSAWS dataset of simulated surgical tasks, and a new clinical dataset of real laparoscopic surgeries. The proposed framework achieves promising results on both datasets, with the state-of-the-art on the simulated dataset advanced from 0.71 Spearman's correlation to 0.80. It is also shown that combining multiple skill aspects yields better performance than relying on a single aspect.

SIENet: Spatial Information Enhancement Network for 3D Object Detection from Point Cloud

LiDAR-based 3D object detection pushes forward an immense influence on autonomous vehicles. Due to the limitation of the intrinsic properties of LiDAR, fewer points are collected at the objects farther away from the sensor. This imbalanced density of point clouds degrades the detection accuracy but is generally neglected by previous works. To address the challenge, we propose a novel two-stage 3D object detection framework, named SIENet. Specifically, we design the Spatial Information Enhancement (SIE) module to predict the spatial shapes of the foreground points within proposals, and extract the structure information to learn the representative features for further box refinement. The predicted spatial shapes are complete and dense point sets, thus the extracted structure information contains more semantic representation. Besides, we design the Hybrid-Paradigm Region Proposal Network (HP-RPN) which includes multiple branches to learn discriminate features and generate accurate proposals for the SIE module. Extensive experiments on the KITTI 3D object detection benchmark show that our elaborately designed SIENet outperforms the state-of-the-art methods by a large margin.

SRDTI: Deep learning-based super-resolution for diffusion tensor MRI

High-resolution diffusion tensor imaging (DTI) is beneficial for probing tissue microstructure in fine neuroanatomical structures, but long scan times and limited signal-to-noise ratio pose significant barriers to acquiring DTI at sub-millimeter resolution. To address this challenge, we propose a deep learning-based super-resolution method entitled "SRDTI" to synthesize high-resolution diffusion-weighted images (DWIs) from low-resolution DWIs. SRDTI employs a deep convolutional neural network (CNN), residual learning and multi-contrast imaging, and generates high-quality results with rich textural details and microstructural information, which are more similar to high-resolution ground truth than those from trilinear and cubic spline interpolation.