Abstract
This thesis studies advanced probabilistic models, including both their theoretical foundations and practical applications, for different semi-supervised learning (SSL) tasks. The proposed probabilistic methods are able to improve the safety of AI systems in real applications by providing reliable uncertainty estimates quickly, and at the same time, achieve competitive performance compared to their deterministic counterparts. The experimental results indicate that the methods proposed in the thesis have great value in safety-critical areas, such as the autonomous driving or medical imaging analysis domain, and pave the way for the future discovery of highly effective and efficient probabilistic approaches in the SSL sector.
Keyword: medical visualization
There is no result
Keyword: interactive volume
There is no result
Keyword: rendering
LightOctree: Lightweight 3D Spatially-Coherent Indoor Lighting Estimation
Authors: Xuecan Wang, Shibang Xiao, Xiaohui Liang
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
We present a lightweight solution for estimating spatially-coherent indoor lighting from a single RGB image. Previous methods for estimating illumination using volumetric representations have overlooked the sparse distribution of light sources in space, necessitating substantial memory and computational resources for achieving high-quality results. We introduce a unified, voxel octree-based illumination estimation framework to produce 3D spatially-coherent lighting. Additionally, a differentiable voxel octree cone tracing rendering layer is proposed to eliminate regular volumetric representation throughout the entire process and ensure the retention of features across different frequency domains. This reduction significantly decreases spatial usage and required floating-point operations without substantially compromising precision. Experimental results demonstrate that our approach achieves high-quality coherent estimation with minimal cost compared to previous methods.
RaSim: A Range-aware High-fidelity RGB-D Data Simulation Pipeline for Real-world Applications
Abstract
In robotic vision, a de-facto paradigm is to learn in simulated environments and then transfer to real-world applications, which poses an essential challenge in bridging the sim-to-real domain gap. While mainstream works tackle this problem in the RGB domain, we focus on depth data synthesis and develop a range-aware RGB-D data simulation pipeline (RaSim). In particular, high-fidelity depth data is generated by imitating the imaging principle of real-world sensors. A range-aware rendering strategy is further introduced to enrich data diversity. Extensive experiments show that models trained with RaSim can be directly applied to real-world scenarios without any finetuning and excel at downstream RGB-D perception tasks.
MM-Gaussian: 3D Gaussian-based Multi-modal Fusion for Localization and Reconstruction in Unbounded Scenes
Abstract
Localization and mapping are critical tasks for various applications such as autonomous vehicles and robotics. The challenges posed by outdoor environments present particular complexities due to their unbounded characteristics. In this work, we present MM-Gaussian, a LiDAR-camera multi-modal fusion system for localization and mapping in unbounded scenes. Our approach is inspired by the recently developed 3D Gaussians, which demonstrate remarkable capabilities in achieving high rendering quality and fast rendering speed. Specifically, our system fully utilizes the geometric structure information provided by solid-state LiDAR to address the problem of inaccurate depth encountered when relying solely on visual solutions in unbounded, outdoor scenarios. Additionally, we utilize 3D Gaussian point clouds, with the assistance of pixel-level gradient descent, to fully exploit the color information in photos, thereby achieving realistic rendering effects. To further bolster the robustness of our system, we designed a relocalization module, which assists in returning to the correct trajectory in the event of a localization failure. Experiments conducted in multiple scenarios demonstrate the effectiveness of our method.
3D Facial Expressions through Analysis-by-Neural-Synthesis
Authors: George Retsinas, Panagiotis P. Filntisis, Radek Danecek, Victoria F. Abrevaya, Anastasios Roussos, Timo Bolkart, Petros Maragos
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
While existing methods for 3D face reconstruction from in-the-wild images excel at recovering the overall face shape, they commonly miss subtle, extreme, asymmetric, or rarely observed expressions. We improve upon these methods with SMIRK (Spatial Modeling for Image-based Reconstruction of Kinesics), which faithfully reconstructs expressive 3D faces from images. We identify two key limitations in existing methods: shortcomings in their self-supervised training formulation, and a lack of expression diversity in the training images. For training, most methods employ differentiable rendering to compare a predicted face mesh with the input image, along with a plethora of additional loss functions. This differentiable rendering loss not only has to provide supervision to optimize for 3D face geometry, camera, albedo, and lighting, which is an ill-posed optimization problem, but the domain gap between rendering and input image further hinders the learning process. Instead, SMIRK replaces the differentiable rendering with a neural rendering module that, given the rendered predicted mesh geometry, and sparsely sampled pixels of the input image, generates a face image. As the neural rendering gets color information from sampled image pixels, supervising with neural rendering-based reconstruction loss can focus solely on the geometry. Further, it enables us to generate images of the input identity with varying expressions while training. These are then utilized as input to the reconstruction model and used as supervision with ground truth geometry. This effectively augments the training data and enhances the generalization for diverse expressions. Our qualitative, quantitative and particularly our perceptual evaluations demonstrate that SMIRK achieves the new state-of-the art performance on accurate expression reconstruction. Project webpage: https://georgeretsi.github.io/smirk/.
Robust Gaussian Splatting
Authors: François Darmon, Lorenzo Porzi, Samuel Rota-Bulò, Peter Kontschieder
Subjects: Computer Vision and Pattern Recognition (cs.CV)
Abstract
In this paper, we address common error sources for 3D Gaussian Splatting (3DGS) including blur, imperfect camera poses, and color inconsistencies, with the goal of improving its robustness for practical applications like reconstructions from handheld phone captures. Our main contribution involves modeling motion blur as a Gaussian distribution over camera poses, allowing us to address both camera pose refinement and motion blur correction in a unified way. Additionally, we propose mechanisms for defocus blur compensation and for addressing color in-consistencies caused by ambient light, shadows, or due to camera-related factors like varying white balancing settings. Our proposed solutions integrate in a seamless way with the 3DGS formulation while maintaining its benefits in terms of training efficiency and rendering speed. We experimentally validate our contributions on relevant benchmark datasets including Scannet++ and Deblur-NeRF, obtaining state-of-the-art results and thus consistent improvements over relevant baselines.
Keyword: cinematic rendering
There is no result
Keyword: volume data
There is no result
Keyword: remote visualization
There is no result
Keyword: direct volume rendering
There is no result
Keyword: mobile device
There is no result
Keyword: transfer function
There is no result
Keyword: retrieval
GenQREnsemble: Zero-Shot LLM Ensemble Prompting for Generative Query Reformulation
Authors: Kaustubh Dhole, Eugene Agichtein
Subjects: Information Retrieval (cs.IR); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
Abstract
Query Reformulation(QR) is a set of techniques used to transform a user's original search query to a text that better aligns with the user's intent and improves their search experience. Recently, zero-shot QR has been shown to be a promising approach due to its ability to exploit knowledge inherent in large language models. By taking inspiration from the success of ensemble prompting strategies which have benefited many tasks, we investigate if they can help improve query reformulation. In this context, we propose an ensemble based prompting technique, GenQREnsemble which leverages paraphrases of a zero-shot instruction to generate multiple sets of keywords ultimately improving retrieval performance. We further introduce its post-retrieval variant, GenQREnsembleRF to incorporate pseudo relevant feedback. On evaluations over four IR benchmarks, we find that GenQREnsemble generates better reformulations with relative nDCG@10 improvements up to 18% and MAP improvements upto 24% over the previous zero-shot state-of-art. On the MSMarco Passage Ranking task, GenQREnsembleRF shows relative gains of 5% MRR using pseudo-relevance feedback, and 9% nDCG@10 using relevant feedback documents.
Uniform Memory Retrieval with Larger Capacity for Modern Hopfield Models
Authors: Dennis Wu, Jerry Yao-Chieh Hu, Teng-Yun Hsiao, Han Liu
Abstract
We propose a two-stage memory retrieval dynamics for modern Hopfield models, termed $\mathtt{U\text{-}Hop}$, with enhanced memory capacity. Our key contribution is a learnable feature map $\Phi$ which transforms the Hopfield energy function into a kernel space. This transformation ensures convergence between the local minima of energy and the fixed points of retrieval dynamics within the kernel space. Consequently, the kernel norm induced by $\Phi$ serves as a novel similarity measure. It utilizes the stored memory patterns as learning data to enhance memory capacity across all modern Hopfield models. Specifically, we accomplish this by constructing a separation loss $\mathcal{L}_\Phi$ that separates the local minima of kernelized energy by separating stored memory patterns in kernel space. Methodologically, $\mathtt{U\text{-}Hop}$ memory retrieval process consists of: \textbf{(Stage~I.)} minimizing separation loss for a more uniformed memory (local minimum) distribution, followed by \textbf{(Stage~II.)} standard Hopfield energy minimization for memory retrieval. This results in a significant reduction of possible meta-stable states in the Hopfield energy function, thus enhancing memory capacity by preventing memory confusion. Empirically, with real-world datasets, we demonstrate that $\mathtt{U\text{-}Hop}$ outperforms all existing modern Hopfield models and SOTA similarity measures, achieving substantial improvements in both associative memory retrieval and deep learning tasks.
Outlier-Efficient Hopfield Layers for Large Transformer-Based Models
Authors: Jerry Yao-Chieh Hu, Pei-Hsuan Chang, Robin Luo, Hong-Yu Chen, Weijian Li, Wei-Po Wang, Han Liu
Abstract
We introduce an Outlier-Efficient Modern Hopfield Model (termed $\mathtt{OutEffHop}$) and use it to address the outlier-induced challenge of quantizing gigantic transformer-based models. Our main contribution is a novel associative memory model facilitating \textit{outlier-efficient} associative memory retrievals. Interestingly, this memory model manifests a model-based interpretation of an outlier-efficient attention mechanism ($\text{Softmax}_1$): it is an approximation of the memory retrieval process of $\mathtt{OutEffHop}$. Methodologically, this allows us to debut novel outlier-efficient Hopfield layers a powerful attention alternative with superior post-quantization performance. Theoretically, the Outlier-Efficient Modern Hopfield Model retains and improves the desirable properties of the standard modern Hopfield models, including fixed point convergence and exponential storage capacity. Empirically, we demonstrate the proposed model's efficacy across large-scale transformer-based and Hopfield-based models (including BERT, OPT, ViT and STanHop-Net), benchmarking against state-of-the-art methods including $\mathtt{Clipped_Softmax}$ and $\mathtt{Gated_Attention}$. Notably, $\mathtt{OutEffHop}$ achieves on average $\sim$22+\% reductions in both average kurtosis and maximum infinity norm of model outputs accross 4 models.
Extract, Define, Canonicalize: An LLM-based Framework for Knowledge Graph Construction
Authors: Bowen Zhang, Harold Soh
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Abstract
In this work, we are interested in automated methods for knowledge graph creation (KGC) from input text. Progress on large language models (LLMs) has prompted a series of recent works applying them to KGC, e.g., via zero/few-shot prompting. Despite successes on small domain-specific datasets, these models face difficulties scaling up to text common in many real-world applications. A principal issue is that in prior methods, the KG schema has to be included in the LLM prompt to generate valid triplets; larger and more complex schema easily exceed the LLMs' context window length. To address this problem, we propose a three-phase framework named Extract-Define-Canonicalize (EDC): open information extraction followed by schema definition and post-hoc canonicalization. EDC is flexible in that it can be applied to settings where a pre-defined target schema is available and when it is not; in the latter case, it constructs a schema automatically and applies self-canonicalization. To further improve performance, we introduce a trained component that retrieves schema elements relevant to the input text; this improves the LLMs' extraction performance in a retrieval-augmented generation-like manner. We demonstrate on three KGC benchmarks that EDC is able to extract high-quality triplets without any parameter tuning and with significantly larger schemas compared to prior works.
A Comparison of Methods for Evaluating Generative IR
Abstract
Information retrieval systems increasingly incorporate generative components. For example, in a retrieval augmented generation (RAG) system, a retrieval component might provide a source of ground truth, while a generative component summarizes and augments its responses. In other systems, a large language model (LLM) might directly generate responses without consulting a retrieval component. While there are multiple definitions of generative information retrieval (Gen-IR) systems, in this paper we focus on those systems where the system's response is not drawn from a fixed collection of documents or passages. The response to a query may be entirely new text never. Since traditional IR evaluation methods break down under this model, we explore various methods that extend traditional offline evaluation approaches to the Gen-IR context. Offline IR evaluation traditionally employs paid human assessors, but increasingly LLMs are replacing human assessment, demonstrating capabilities similar or superior to crowdsourced labels. Given that Gen-IR systems do not generate responses from a fixed set, we assume that methods for Gen-IR evaluation must largely depend on LLM-generated labels. Along with methods based on binary and graded relevance, we explore methods based on explicit subtopics, pairwise preferences, and embeddings. We first validate these methods against human assessments on several TREC Deep Learning Track tasks; we then apply these methods to evaluate the output of several purely generative systems. For each method we consider both its ability to act autonomously, without the need for human labels or other input, and its ability to support human auditing. To trust these methods, we must be assured that their results align with human assessments. In order to do so, evaluation criteria must be transparent, so that outcomes can be audited by human assessors.
No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance
Authors: Vishaal Udandarao, Ameya Prabhu, Adhiraj Ghosh, Yash Sharma, Philip H.S. Torr, Adel Bibi, Samuel Albanie, Matthias Bethge
Subjects: Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Machine Learning (cs.LG)
Abstract
Web-crawled pretraining datasets underlie the impressive "zero-shot" evaluation performance of multimodal models, such as CLIP for classification/retrieval and Stable-Diffusion for image generation. However, it is unclear how meaningful the notion of "zero-shot" generalization is for such multimodal models, as it is not known to what extent their pretraining datasets encompass the downstream concepts targeted for during "zero-shot" evaluation. In this work, we ask: How is the performance of multimodal models on downstream concepts influenced by the frequency of these concepts in their pretraining datasets? We comprehensively investigate this question across 34 models and five standard pretraining datasets (CC-3M, CC-12M, YFCC-15M, LAION-400M, LAION-Aesthetics), generating over 300GB of data artifacts. We consistently find that, far from exhibiting "zero-shot" generalization, multimodal models require exponentially more data to achieve linear improvements in downstream "zero-shot" performance, following a sample inefficient log-linear scaling trend. This trend persists even when controlling for sample-level similarity between pretraining and downstream datasets, and testing on purely synthetic data distributions. Furthermore, upon benchmarking models on long-tailed data sampled based on our analysis, we demonstrate that multimodal models across the board perform poorly. We contribute this long-tail test set as the "Let it Wag!" benchmark to further research in this direction. Taken together, our study reveals an exponential need for training data which implies that the key to "zero-shot" generalization capabilities under large-scale training paradigms remains to be found.
Dwell in the Beginning: How Language Models Embed Long Documents for Dense Retrieval
Authors: João Coelho, Bruno Martins, João Magalhães, Jamie Callan, Chenyan Xiong
Subjects: Information Retrieval (cs.IR); Computation and Language (cs.CL)
Abstract
This study investigates the existence of positional biases in Transformer-based models for text representation learning, particularly in the context of web document retrieval. We build on previous research that demonstrated loss of information in the middle of input sequences for causal language models, extending it to the domain of representation learning. We examine positional biases at various stages of training for an encoder-decoder model, including language model pre-training, contrastive pre-training, and contrastive fine-tuning. Experiments with the MS-MARCO document collection reveal that after contrastive pre-training the model already generates embeddings that better capture early contents of the input, with fine-tuning further aggravating this effect.
Keyword: video retrieval
There is no result
Keyword: mobile
A Tutorial on Gaussian Process Learning-based Model Predictive Control
Authors: Jie Wang, Youmin Zhang
Subjects: Robotics (cs.RO); Machine Learning (cs.LG); Systems and Control (eess.SY)
Abstract
This tutorial provides a systematic introduction to Gaussian process learning-based model predictive control (GP-MPC), an advanced approach integrating Gaussian process (GP) with model predictive control (MPC) for enhanced control in complex systems. It begins with GP regression fundamentals, illustrating how it enriches MPC with enhanced predictive accuracy and robust handling of uncertainties. A central contribution of this tutorial is the first detailed, systematic mathematical formulation of GP-MPC in literature, focusing on deriving the approximation of means and variances propagation for GP multi-step predictions. Practical applications in robotics control, such as path-following for mobile robots in challenging terrains and mixed-vehicle platooning, are discussed to demonstrate the real-world effectiveness and adaptability of GP-MPC. This tutorial aims to make GP-MPC accessible to researchers and practitioners, enriching the learning-based control field with in-depth theoretical and practical insights and fostering further innovations in complex system control.
Multi-Task Learning for Lung sound & Lung disease classification
Authors: Suma K V, Deepali Koppad, Preethi Kumar, Neha A Kantikar, Surabhi Ramesh
Abstract
In recent years, advancements in deep learning techniques have considerably enhanced the efficiency and accuracy of medical diagnostics. In this work, a novel approach using multi-task learning (MTL) for the simultaneous classification of lung sounds and lung diseases is proposed. Our proposed model leverages MTL with four different deep learning models such as 2D CNN, ResNet50, MobileNet and Densenet to extract relevant features from the lung sound recordings. The ICBHI 2017 Respiratory Sound Database was employed in the current study. The MTL for MobileNet model performed better than the other models considered, with an accuracy of74\% for lung sound analysis and 91\% for lung diseases classification. Results of the experimentation demonstrate the efficacy of our approach in classifying both lung sounds and lung diseases concurrently. In this study,using the demographic data of the patients from the database, risk level computation for Chronic Obstructive Pulmonary Disease is also carried out. For this computation, three machine learning algorithms namely Logistic Regression, SVM and Random Forest classifierswere employed. Among these ML algorithms, the Random Forest classifier had the highest accuracy of 92\%.This work helps in considerably reducing the physician's burden of not just diagnosing the pathology but also effectively communicating to the patient about the possible causes or outcomes.
Next Generation Multiple Access for IMT Towards 2030 and Beyond
Authors: Zhiguo Ding, Robert Schober, Pingzhi Fan, H. Vincent Poor
Subjects: Information Theory (cs.IT); Signal Processing (eess.SP)
Abstract
Multiple access techniques are fundamental to the design of wireless communication systems, since many crucial components of such systems depend on the choice of the multiple access technique. Because of the importance of multiple access, there has been an ongoing quest during the past decade to develop next generation multiple access (NGMA). Among those potential candidates for NGMA, non-orthogonal multiple access (NOMA) has received significant attention from both the industrial and academic research communities, and has been highlighted in the recently published International Mobile Telecommunications (IMT)-2030 Framework. However, there is still no consensus in the research community about how exactly NOMA assisted NGMA should be designed. This perspective is to outline three important features of NOMA assisted NGMA, namely multi-domain utilization, multi-mode compatibility, and multi-dimensional optimality, where important directions for future research into the design of NOMA assisted NGMA are also discussed.
Bidirectional Human Interactive AI Framework for Social Robot Navigation
Abstract
Trustworthiness is a crucial concept in the context of human-robot interaction. Cooperative robots must be transparent regarding their decision-making process, especially when operating in a human-oriented environment. This paper presents a comprehensive end-to-end framework aimed at fostering trustworthy bidirectional human-robot interaction in collaborative environments for the social navigation of mobile robots. Our method enables a mobile robot to predict the trajectory of people and adjust its route in a socially-aware manner. In case of conflict between human and robot decisions, detected through visual examination, the route is dynamically modified based on human preference while verbal communication is maintained. We present our pipeline, framework design, and preliminary experiments that form the foundation of our proposition.
Machine Learning-Aided Cooperative Localization under Dense Urban Environment
Authors: Hoon Lee, Hong Ki Kim, Seung Hyun Oh, Sang Hyun Lee
Subjects: Information Theory (cs.IT); Signal Processing (eess.SP)
Abstract
Future wireless network technology provides automobiles with the connectivity feature to consolidate the concept of vehicular networks that collaborate on conducting cooperative driving tasks. The full potential of connected vehicles, which promises road safety and quality driving experience, can be leveraged if machine learning models guarantee the robustness in performing core functions including localization and controls. Location awareness, in particular, lends itself to the deployment of location-specific services and the improvement of the operation performance. The localization entails direct communication to the network infrastructure, and the resulting centralized positioning solutions readily become intractable as the network scales up. As an alternative to the centralized solutions, this article addresses decentralized principle of vehicular localization reinforced by machine learning techniques in dense urban environments with frequent inaccessibility to reliable measurement. As such, the collaboration of multiple vehicles enhances the positioning performance of machine learning approaches. A virtual testbed is developed to validate this machine learning model for real-map vehicular networks. Numerical results demonstrate universal feasibility of cooperative localization, in particular, for dense urban area configurations.
player2vec: A Language Modeling Approach to Understand Player Behavior in Games
Abstract
Methods for learning latent user representations from historical behavior logs have gained traction for recommendation tasks in e-commerce, content streaming, and other settings. However, this area still remains relatively underexplored in video and mobile gaming contexts. In this work, we present a novel method for overcoming this limitation by extending a long-range Transformer model from the natural language processing domain to player behavior data. We discuss specifics of behavior tracking in games and propose preprocessing and tokenization approaches by viewing in-game events in an analogous way to words in sentences, thus enabling learning player representations in a self-supervised manner in the absence of ground-truth annotations. We experimentally demonstrate the efficacy of the proposed approach in fitting the distribution of behavior events by evaluating intrinsic language modeling metrics. Furthermore, we qualitatively analyze the emerging structure of the learned embedding space and show its value for generating insights into behavior patterns to inform downstream applications.
Keyword: volume render
There is no result
Keyword: volumetric render
There is no result
Keyword: remote render
There is no result
Keyword: hybrid render
There is no result
Keyword: raycast
There is no result
Keyword: medical imaging
Exploring Probabilistic Models for Semi-supervised Learning
Keyword: medical visualization
There is no result
Keyword: interactive volume
There is no result
Keyword: rendering
LightOctree: Lightweight 3D Spatially-Coherent Indoor Lighting Estimation
RaSim: A Range-aware High-fidelity RGB-D Data Simulation Pipeline for Real-world Applications
MM-Gaussian: 3D Gaussian-based Multi-modal Fusion for Localization and Reconstruction in Unbounded Scenes
3D Facial Expressions through Analysis-by-Neural-Synthesis
Robust Gaussian Splatting
Keyword: cinematic rendering
There is no result
Keyword: volume data
There is no result
Keyword: remote visualization
There is no result
Keyword: direct volume rendering
There is no result
Keyword: mobile device
There is no result
Keyword: transfer function
There is no result
Keyword: retrieval
GenQREnsemble: Zero-Shot LLM Ensemble Prompting for Generative Query Reformulation
Uniform Memory Retrieval with Larger Capacity for Modern Hopfield Models
Outlier-Efficient Hopfield Layers for Large Transformer-Based Models
Extract, Define, Canonicalize: An LLM-based Framework for Knowledge Graph Construction
A Comparison of Methods for Evaluating Generative IR
No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance
Dwell in the Beginning: How Language Models Embed Long Documents for Dense Retrieval
Keyword: video retrieval
There is no result
Keyword: mobile
A Tutorial on Gaussian Process Learning-based Model Predictive Control
Multi-Task Learning for Lung sound & Lung disease classification
Next Generation Multiple Access for IMT Towards 2030 and Beyond
Bidirectional Human Interactive AI Framework for Social Robot Navigation
Machine Learning-Aided Cooperative Localization under Dense Urban Environment
player2vec: A Language Modeling Approach to Understand Player Behavior in Games
Keyword: smartphone
There is no result
Keyword: medical volume data
There is no result