arXiv subjects

Yize Zhao

Publications and source records attributed to Yize Zhao.

At least 19 recordsLinked to original sources

arXiv2026

Robust Community Detection for Noisy Networks with Covariates: Application to Functional Brain Networks

Community detection is fundamental to understanding the modular organization in functional brain networks, yet noise in neuroimaging-derived networks and auxiliary node-level covariates pose critical challenges. Existing methods typically either assume networks are noise-free or ignore covariate information. We propose a Bayesian framework for recovering a shared latent community structure from multiple noisy network realizations and auxiliary covariates. The model combines a degree-corrected stochastic block model for the latent network, a block-structured noise model linking noisy observations to latent edges, and a covariate cluster model for node-level attributes. This specification allows anatomical or functional attributes of regions of interest to contribute information when network signals are weak or sparse. We develop an efficient Markov chain Monte Carlo algorithm for posterior sampling and select the number of communities using the widely applicable information criterion, avoiding prior specification of this quantity. Simulation studies demonstrate improved community recovery relative to existing methods across varying noise levels, covariate signal strengths, and numbers of noisy networks, with larger gains when network noise is moderate to high or only a small number of noisy networks is available. Applications to functional brain networks from the Alzheimer's Disease Neuroimaging Initiative and the Human Connectome Project identify biologically interpretable structures and capture disease-related reorganization and individual-level variation.

stat.ME↗

arXiv2026

Structure Before Collapse: Transient semantic geometry in next-token prediction

Neural Collapse predicts that balanced one-hot classification pushes model representations to be equally far from each other; a symmetric configuration that depends only on the output label and ignores any semantic similarity in the inputs. This creates a puzzle: next-token prediction language models are trained predominantly (as context length increases) with one-hot labels: the same context is very unlikely to appear twice in training with different labels. However, they clearly learn latent structural features. That is, despite the one-hot training regime, a language model's contextual embeddings represent the fact that the next word in ''Mary broke the ___'' is likely to be filled by tokens in the latent classes of a) medium-sized, b) rigid, c) inanimate nouns. How does gradient descent find such categorical semantic structure when co-occurrence statistics collapse to one-hot sparsity, eliminating any shared next-tokens among different contexts? To investigate this tension we identify three synthetic controlled settings where inputs have latent semantic factors but are mapped to distinct one-hot labels. We find that semantic geometry emerges early in training, and that representations cluster by shared attributes despite receiving no explicit supervision to do so. This structure is transient: with sufficient capacity and time, the model eventually reaches the predicted symmetric state where all representations are equally separated. We study this phase transition through Gram matrix analysis and propose a preliminary modification to the commonly used unconstrained features model to capture the emergent semantic geometry.

cs.LG↗

arXiv2026

iLENS: Interpretable LLM-Guided Mixture-of-Experts for Neuroimaging Survival Analysis

Alzheimer's Disease (AD) is a complex neurodegenerative disorder that continues to impact millions of people worldwide. Predicting AD conversion during the prodromal stage remains critical for disease understanding and patient care. As such, survival models are widely used for AD risk prediction, yet they are typically static predictors with limited interpretability and no capacity for natural language reasoning. In this work, we propose iLENS, an interpretable large language model (LLM) guided framework based on mixture-of-experts (MoE) for survival prediction in AD conversion. Our approach uses LLM to synthesize structured neuroimaging measurements and unstructured information to guide expert routing. Our framework demonstrates competitive predictive performance and capability in patient subtyping. Furthermore, our framework provides transparent, biologically grounded rationales for its routing decisions, bridging the gap between high-performance survival analysis and interpretable clinical decision support.

cs.LG↗

arXiv2026

Disentangling Latent Risk Pathways via Bayesian Hypergraph Inference

Electronic health records (EHR) pose large-scale multi-disease modeling problems in which many outcomes are rare and strongly influenced by shared risk factors. While modern approaches achieve strong predictive performance, they often treat diseases independently or rely on black-box architectures, offering limited insight into how risk factors organize disease risk and little principled uncertainty quantification. We introduce a Bayesian hypergraph inference framework that reframes multi-disease modeling around latent, risk-factor-modulated disease pathways. Risk factors act on hyperedges, latent disease subsets with shared risk patterns, allowing diseases to participate in multiple distinct pathways and enabling interpretable, higher-order structure beyond pairwise associations. A repulsion prior encourages parsimonious and identifiable structure, while posterior inference provides calibrated uncertainty over both disease groupings and risk-factor influence. To enable scalable inference on large EHR datasets, we develop a structured variational inference algorithm that preserves logical dependencies among hyperedge existence, disease membership, and pathway-level effects. Experiments on simulated data and UK Biobank demonstrate stable and interpretable disease pathway structure, well-calibrated uncertainty, improved estimation for rare diseases, and competitive predictive performance.

stat.ML↗

arXiv2026

Workflow Closure Is Not Scientific Closure in Auto-Research Systems

This paper argues that workflow closure is not scientific closure in auto-research systems. Current systems can increasingly complete research-like loops internally, moving from idea generation to experiment execution, writing, and self-evaluation. That achievement is real, but it does not by itself give the resulting outputs scientific standing. We argue that trustworthy auto-research should not aim for autonomous self-sufficiency, but should aim for autonomous execution under non-autonomous epistemic control. Based on a survey of more than 100 recent papers and repositories in this rapidly emerging area, together with a structured audit of 21 representative systems, we diagnose a recurring and structurally connected failure pattern: objective collapse, in which single-proxy targets replace multi-objective scientific aims; validation collapse, in which internal self-evaluation replaces independent validation; and acceptance collapse, in which benchmark scores or publication-shaped artifacts replace mechanisms for domain-level critique, reuse, and integration. These collapses are not inherent limits of autonomy but correctable design choices. Accordingly, we outline potential remedies across objective signal, validation, and output pathway to spark community discussion.

cs.SE↗

arXiv2026

Why Loss Re-weighting Works If You Stop Early: Training Dynamics of Unconstrained Features

The application of loss reweighting in modern deep learning presents a nuanced picture. While it fails to alter the terminal learning phase in overparameterized deep neural networks (DNNs) trained on high-dimensional datasets, empirical evidence consistently shows it offers significant benefits early in training. To transparently demonstrate and analyze this phenomenon, we introduce a small-scale model (SSM). This model is specifically designed to abstract the inherent complexities of both the DNN architecture and the input data, while maintaining key information about the structure of imbalance within its spectral components. On the one hand, the SSM reveals how vanilla empirical risk minimization preferentially learns to distinguish majority classes over minorities early in training, consequently delaying minority learning. In stark contrast, reweighting restores balanced learning dynamics, enabling the simultaneous learning of features associated with both majorities and minorities.

cs.LG↗

arXiv2025

BlinkBud: Detecting Hazards from Behind via Sampled Monocular 3D Detection on a Single Earbud

Failing to be aware of speeding vehicles approaching from behind poses a huge threat to the road safety of pedestrians and cyclists. In this paper, we propose BlinkBud, which utilizes a single earbud and a paired phone to online detect hazardous objects approaching from behind of a user. The core idea is to accurately track visually identified objects utilizing a small number of sampled camera images taken from the earbud. To minimize the power consumption of the earbud and the phone while guaranteeing the best tracking accuracy, a novel 3D object tracking algorithm is devised, integrating both a Kalman filter based trajectory estimation scheme and an optimal image sampling strategy based on reinforcement learning. Moreover, the impact of constant user head movements on the tracking accuracy is significantly eliminated by leveraging the estimated pitch and yaw angles to correct the object depth estimation and align the camera coordinate system to the user's body coordinate system, respectively. We implement a prototype BlinkBud system and conduct extensive real-world experiments. Results show that BlinkBud is lightweight with ultra-low mean power consumptions of 29.8 mW and 702.6 mW on the earbud and smartphone, respectively, and can accurately detect hazards with a low average false positive ratio (FPR) and false negative ratio (FNR) of 4.90% and 1.47%, respectively.

cs.CV↗

arXiv2025

How Muon's Spectral Design Benefits Generalization: A Study on Imbalanced Data

The growing adoption of spectrum-aware matrix-valued optimizers such as Muon and Shampoo in deep learning motivates a systematic study of their generalization properties and, in particular, when they might outperform competitive algorithms. We approach this question by introducing appropriate simplifying abstractions as follows: First, we use imbalanced data as a testbed. Second, we study the canonical form of such optimizers, which is Spectral Gradient Descent (SpecGD) -- each update step is $UV^T$ where $U\Sigma V^T$ is the truncated SVD of the gradient. Third, within this framework we identify a canonical setting for which we precisely quantify when SpecGD outperforms vanilla Euclidean GD. For a Gaussian mixture data model and both linear and bilinear models, we show that unlike GD, which prioritizes learning dominant principal components of the data first, SpecGD learns all principal components of the data at equal rates. We demonstrate how this translates to a growing gap in class balanced loss favoring SpecGD early in training and further show that the gap remains consistent even when the GD counterpart uses adaptive step-sizes via normalization. By extending the analysis to deep linear models, we show that depth amplifies these effects. We empirically verify our theoretical findings on a variety of imbalanced datasets. Our experiments compare practical variants of spectral methods, like Muon and Shampoo, against their Euclidean counterparts and Adam. The results validate our findings that these spectral optimizers achieve superior generalization by promoting a more balanced learning of the data's underlying components.

cs.LG↗

arXiv2025

Geometry of Semantics in Next-Token Prediction: How Optimization Implicitly Organizes Linguistic Representations

We investigate how next-token prediction (NTP) optimization leads language models to extract and organize semantic structure from text. Our analysis, based on a tractable mathematical model and controlled synthetic data, reveals that NTP implicitly guides models to factor a centered support matrix encoding context-to-next-token co-occurrence patterns via singular value decomposition (SVD). While models never explicitly construct this matrix, learned word and context embeddings converge to its SVD factors, with singular vectors encoding latent semantic concepts through their sign patterns. We demonstrate that concepts corresponding to larger singular values are learned earlier during training, yielding a natural semantic hierarchy where broad categories emerge before fine-grained ones. This insight motivates orthant-based clustering, a method that combines concept signs to identify interpretable semantic categories. We validate our findings on synthetic datasets and pretrained language models, recovering diverse semantic structures such as grammatical categories, named entity types, and topical distinctions (medical, entertainment). Our work bridges classical distributional semantics and neural collapse geometry, characterizing how gradient-based optimization implicitly determines both the matrix representation and factorization method that encode semantic structure.

cs.CL↗

arXiv2024

Cost efficiency of fMRI studies using resting-state vs task-based functional connectivity

We investigate whether and how we can improve the cost efficiency of neuroimaging studies with well-tailored fMRI tasks. The comparative study is conducted using a novel network science-driven Bayesian connectome-based predictive method, which incorporates network theories in model building and substantially improves precision and robustness in imaging biomarker detection. The robustness of the method lays the foundation for identifying predictive power differential across fMRI task conditions if such difference exists. When applied to a clinically heterogeneous transdiagnostic cohort, we found shared and distinct functional fingerprints of neuropsychological outcomes across seven fMRI conditions. For example, emotional N-back memory task was found to be less optimal for negative emotion outcomes, and gradual-onset continuous performance task was found to have stronger links with sensitivity and sociability outcomes than with cognitive control outcomes. Together, our results show that there are unique optimal pairings of task-based fMRI conditions and neuropsychological outcomes that should not be ignored when designing well-powered neuroimaging studies.

stat.AP↗

arXiv2024

DARE the Extreme: Revisiting Delta-Parameter Pruning For Fine-Tuned Models

Storing open-source fine-tuned models separately introduces redundancy and increases response times in applications utilizing multiple models. Delta-parameter pruning (DPP), particularly the random drop and rescale (DARE) method proposed by Yu et al., addresses this by pruning the majority of delta parameters--the differences between fine-tuned and pre-trained model weights--while typically maintaining minimal performance loss. However, DARE fails when either the pruning rate or the magnitude of the delta parameters is large. We highlight two key reasons for this failure: (1) an excessively large rescaling factor as pruning rates increase, and (2) high mean and variance in the delta parameters. To push DARE's limits, we introduce DAREx (DARE the eXtreme), which features two algorithmic improvements: (1) DAREx-q, a rescaling factor modification that significantly boosts performance at high pruning rates (e.g., >30 % on COLA and SST2 for encoder models, with even greater gains in decoder models), and (2) DAREx-L2, which combines DARE with AdamR, an in-training method that applies appropriate delta regularization before DPP. We also demonstrate that DAREx-q can be seamlessly combined with vanilla parameter-efficient fine-tuning techniques like LoRA and can facilitate structural DPP. Additionally, we revisit the application of importance-based pruning techniques within DPP, demonstrating that they outperform random-based methods when delta parameters are large. Through this comprehensive study, we develop a pipeline for selecting the most appropriate DPP method under various practical scenarios.

cs.LG↗

arXiv2024

Implicit Geometry of Next-token Prediction: From Language Sparsity Patterns to Model Representations

Next-token prediction (NTP) over large text corpora has become the go-to paradigm to train large language models. Yet, it remains unclear how NTP influences the mapping of linguistic patterns to geometric properties of the resulting model representations. We frame training of large language models as soft-label classification over sparse probabilistic label vectors, coupled with an analytical approximation that allows unrestricted generation of context embeddings. This approach links NTP training to rank-constrained, nuclear-norm regularized optimization in the logit domain, offering a framework for analyzing the geometry of word and context embeddings. In large embedding spaces, we find that NTP implicitly favors learning logits with a sparse plus low-rank structure. While the sparse component captures the co-occurrence frequency of context-word pairs, the orthogonal low-rank component, which becomes dominant as training progresses, depends solely on the sparsity pattern of the co-occurrence matrix. Consequently, when projected onto an appropriate subspace, representations of contexts that are followed by the same set of next-tokens collapse, a phenomenon we term subspace-collapse. We validate our findings on synthetic and small-scale real language datasets. Finally, we outline potential research directions aimed at deepening the understanding of NTP's influence on the learning of linguistic patterns and regularities.

cs.CL↗

arXiv2024

Bayesian thresholded modeling for integrating brain node and network predictors

Progress in neuroscience has provided unprecedented opportunities to advance our understanding of brain alterations and their correspondence to phenotypic profiles. With data collected from various imaging techniques, studies have integrated different types of information ranging from brain structure, function, or metabolism. More recently, an emerging way to categorize imaging traits is through a metric hierarchy, including localized node-level measurements and interactive network-level metrics. However, limited research has been conducted to integrate these different hierarchies and achieve a better understanding of the neurobiological mechanisms and communications. In this work, we address this literature gap by proposing a Bayesian regression model under both vector-variate and matrix-variate predictors. To characterize the interplay between different predicting components, we propose a set of biologically plausible prior models centered on an innovative joint thresholded prior. This captures the coupling and grouping effect of signal patterns, as well as their spatial contiguity across brain anatomy. By developing a posterior inference, we can identify and quantify the uncertainty of signaling node- and network-level neuromarkers, as well as their predictive mechanism for phenotypic outcomes. Through extensive simulations, we demonstrate that our proposed method outperforms the alternative approaches substantially in both out-of-sample prediction and feature selection. By implementing the model to study children's general mental abilities, we establish a powerful predictive mechanism based on the identified task contrast traits and resting-state sub-networks.

stat.ME↗

arXiv2024

Supervised brain node and network construction under voxel-level functional imaging

Recent advancements in understanding the brain's functional organization related to behavior have been pivotal, particularly in the development of predictive models based on brain connectivity. Traditional methods in this domain often involve a two-step process by first constructing a connectivity matrix from predefined brain regions, and then linking these connections to behaviors or clinical outcomes. However, these approaches with unsupervised node partitions predict outcomes inefficiently with independently established connectivity. In this paper, we introduce the Supervised Brain Parcellation (SBP), a brain node parcellation scheme informed by the downstream predictive task. With voxel-level functional time courses generated under resting-state or cognitive tasks as input, our approach clusters voxels into nodes in a manner that maximizes the correlation between inter-node connections and the behavioral outcome, while also accommodating intra-node homogeneity. We rigorously evaluate the SBP approach using resting-state and task-based fMRI data from both the Adolescent Brain Cognitive Development (ABCD) study and the Human Connectome Project (HCP). Our analyses show that SBP significantly improves out-of-sample connectome-based predictive performance compared to conventional step-wise methods under various brain atlases. This advancement holds promise for enhancing our understanding of brain functional architectures with behavior and establishing more informative network neuromarkers for clinical applications.

stat.AP↗

arXiv2024

Volume-optimal persistence homological scaffolds of hemodynamic networks covary with MEG theta-alpha aperiodic dynamics

Higher-order properties of functional magnetic resonance imaging (fMRI) induced connectivity have been shown to unravel many exclusive topological and dynamical insights beyond pairwise interactions. Nonetheless, whether these fMRI-induced higher-order properties play a role in disentangling other neuroimaging modalities' insights remains largely unexplored and poorly understood. In this work, by analyzing fMRI data from the Human Connectome Project Young Adult dataset using persistent homology, we discovered that the volume-optimal persistence homological scaffolds of fMRI-based functional connectomes exhibited conservative topological reconfigurations from the resting state to attentional task-positive state. Specifically, while reflecting the extent to which each cortical region contributed to functional cycles following different cognitive demands, these reconfigurations were constrained such that the spatial distribution of cavities in the connectome is relatively conserved. Most importantly, such level of contributions covaried with powers of aperiodic activities mostly within the theta-alpha (4-12 Hz) band measured by magnetoencephalography (MEG). This comprehensive result suggests that fMRI-induced hemodynamics and MEG theta-alpha aperiodic activities are governed by the same functional constraints specific to each cortical morpho-structure. Methodologically, our work paves the way toward an innovative computing paradigm in multimodal neuroimaging topological learning.

q-bio.NC↗

arXiv2024

A principled framework to assess the information-theoretic fitness of brain functional sub-circuits

In systems and network neuroscience, many common practices in brain connectomic analysis are often not properly scrutinized. One such practice is mapping a predetermined set of sub-circuits, like functional networks (FNs), onto subjects' functional connectomes (FCs) without adequately assessing the information-theoretic appropriateness of the partition. Another practice that goes unchallenged is thresholding weighted FCs to remove spurious connections without justifying the chosen threshold. This paper leverages recent theoretical advances in Stochastic Block Models (SBMs) to formally define and quantify the information-theoretic fitness (e.g., prominence) of a predetermined set of FNs when mapped to individual FCs under different fMRI task conditions. Our framework allows for evaluating any combination of FC granularity, FN partition, and thresholding strategy, thereby optimizing these choices to preserve important topological features of the human brain connectomes. By applying to the Human Connectome Project with Schaefer parcellations at multiple levels of granularity, the framework showed that the common thresholding value of 0.25 was indeed information-theoretically valid for group-average FCs despite its previous lack of justification. Our results pave the way for the proper use of FNs and thresholding methods and provide insights for future research in individualized parcellations.

q-bio.NC↗

arXiv2024

Learnable Community-Aware Transformer for Brain Connectome Analysis with Token Clustering

Neuroscientific research has revealed that the complex brain network can be organized into distinct functional communities, each characterized by a cohesive group of regions of interest (ROIs) with strong interconnections. These communities play a crucial role in comprehending the functional organization of the brain and its implications for neurological conditions, including Autism Spectrum Disorder (ASD) and biological differences, such as in gender. Traditional models have been constrained by the necessity of predefined community clusters, limiting their flexibility and adaptability in deciphering the brain's functional organization. Furthermore, these models were restricted by a fixed number of communities, hindering their ability to accurately represent the brain's dynamic nature. In this study, we present a token clustering brain transformer-based model ($\texttt{TC-BrainTF}$) for joint community clustering and classification. Our approach proposes a novel token clustering (TC) module based on the transformer architecture, which utilizes learnable prompt tokens with orthogonal loss where each ROI embedding is projected onto the prompt embedding space, effectively clustering ROIs into communities and reducing the dimensions of the node representation via merging with communities. Our results demonstrate that our learnable community-aware model $\texttt{TC-BrainTF}$ offers improved accuracy in identifying ASD and classifying genders through rigorous testing on ABIDE and HCP datasets. Additionally, the qualitative analysis on $\texttt{TC-BrainTF}$ has demonstrated the effectiveness of the designed TC module and its relevance to neuroscience interpretations.

q-bio.NC↗

arXiv2023

Learning High-Order Relationships of Brain Regions

Discovering reliable and informative relationships among brain regions from functional magnetic resonance imaging (fMRI) signals is essential in phenotypic predictions. Most of the current methods fail to accurately characterize those interactions because they only focus on pairwise connections and overlook the high-order relationships of brain regions. We propose that these high-order relationships should be maximally informative and minimally redundant (MIMR). However, identifying such high-order relationships is challenging and under-explored due to the exponential search space and the absence of a tractable objective. In response to this gap, we propose a novel method named HYBRID which aims to extract MIMR high-order relationships from fMRI data. HYBRID employs a CONSTRUCTOR to identify hyperedge structures, and a WEIGHTER to compute a weight for each hyperedge, which avoids searching in exponential space. HYBRID achieves the MIMR objective through an innovative information bottleneck framework named multi-head drop-bottleneck with theoretical guarantees. Our comprehensive experiments demonstrate the effectiveness of our model. Our model outperforms the state-of-the-art predictive model by an average of 11.2%, regarding the quality of hyperedges measured by CPM, a standard protocol for studying brain connections.

q-bio.NC↗