arXiv subjects

Yoshihiro Nagano

Publications and source records attributed to Yoshihiro Nagano.

11 recordsLinked to original sources

arXiv2026

Overcoming Output Dimension Collapse: When Sparsity Enables Zero-shot Brain-to-Image Reconstruction at Small Data Scales

Advances in brain-to-image reconstruction are enabling us to externalize the subjective visual experiences encoded in the brain as images. A key challenge in this task is data scarcity: a translator that maps brain activity to latent image features is trained on a limited number of brain-image pairs, making the translator a bottleneck for zero-shot reconstruction beyond the training stimuli. In this paper, we mathematically analyze the behavior of two translators commonly used in recent reconstruction pipelines: naive multivariate linear regression and sparse multivariate linear regression. We define the data scale as the ratio of the number of training samples to the latent feature dimensionality and characterize the behavior of each model across data scales. Building on a standard structural property of naive multivariate regression, we first show that the resulting ``output dimension collapse'' can become a practical generalization bottleneck in brain-to-image reconstruction. We introduce the best prediction diagnostic, which is computable without brain activity, to quantify the practical impact of this collapse. We then analyze sparse linear regression models in a student--teacher framework and derive expressions for the prediction error in terms of data scale and other sparsity-related parameters. Our analysis clarifies when variable selection can reduce prediction error at small data scales by exploiting the sparsity of the brain-to-feature mapping. Our findings provide quantitative guidelines for diagnosing output dimension collapse and for designing effective translators and feature representations for zero-shot reconstruction.

q-bio.NC↗

arXiv2025

Advancing credibility and transparency in brain-to-image reconstruction research: Reanalysis of Koide-Majima, Nishimoto, and Majima (Neural Networks, 2024)

A recent high-profile study by Koide-Majima et al. (2024) claimed a major advance in reconstructing visual imagery from brain activity using a novel variant of a generative AI-based method. However, our independent reanalysis reveals multiple methodological concerns that raise questions about the validity of their conclusions. Specifically, our evaluation demonstrates that: (1) the reconstruction results are biased by selective reporting of only the best-performing examples at multiple levels; (2) performance is artificially inflated by circular metrics that fail to reflect perceptual accuracy; (3) fair baseline comparisons reveal no discernible advantages of the study's key innovations over existing techniques; (4) the central "Bayesian" sampling component is functionally inert, producing outcomes identical to the standard optimization result; and (5) even if the component were successfully implemented, the claims of Bayesian novelty are unsubstantiated, as the proposed method does not leverage the principles of a proper Bayesian framework. These systemic issues necessitate a critical reassessment of the study's contributions. This commentary dissects these deficiencies to underscore the need for greater credibility and transparency in the rapidly advancing field of brain decoding.

q-bio.NC↗

arXiv2025

Readout Representation: Redefining Neural Codes by Input Recovery

Sensory representation is typically understood through a hierarchical-causal framework where progressively abstract features are extracted sequentially. However, this causal view fails to explain misrepresentation, a phenomenon better handled by an informational view based on decodable content. This creates a tension: how does a system that abstracts away details still preserve the fine-grained information needed for downstream functions? We propose readout representation to resolve this, defining representation by the information recoverable from features rather than their causal origin. Empirically, we show that inputs can be accurately reconstructed even from heavily perturbed mid-level features, demonstrating that a single input corresponds to a broad, redundant region of feature space, challenging the causal mapping perspective. To quantify this property, we introduce representation size, a metric linked to model robustness and representational redundancy. Our framework offers a new lens for analyzing how both biological and artificial neural systems learn complex features while maintaining robust, information-rich representations of the world.

q-bio.NC↗

arXiv2025

Spurious reconstruction from brain activity

Advances in brain decoding, particularly visual image reconstruction, have sparked discussions about the societal implications and ethical considerations of neurotechnology. As these methods aim to recover visual experiences from brain activity and achieve prediction beyond training samples (zero-shot prediction), it is crucial to assess their capabilities and limitations to inform public expectations and regulations. Our case study of recent text-guided reconstruction methods, which leverage a large-scale dataset (Natural Scene Dataset, NSD) and text-to-image diffusion models, reveals limitations in their generalizability. We found poor performance when applying these methods to a different dataset designed to prevent category overlaps between training and test sets. UMAP visualization of the text features with NSD images showed a limited diversity of semantic and visual clusters, with overlap between training and test sets. Formal analysis and simulations demonstrated that clustered training samples can lead to "output dimension collapse," restricting predictable output feature dimensions. Simulations further showed that diversifying the training set improved generalizability. However, text features alone are insufficient for mapping to the visual space. We argue that recent realistic reconstructions may primarily be a blend of classification into trained categories and generation of inauthentic images through text-to-image diffusion (hallucination). Diverse datasets and compositional representations spanning the image space are essential for genuine zero-shot prediction. Interdisciplinary discussions grounded in understanding the current capabilities and limitations, as well as ethical considerations, of the technology are crucial for its responsible development.

q-bio.NC↗

arXiv2025

Exploring internal representation of self-supervised networks: few-shot learning abilities and comparison with human semantics and recognition of objects

Recent advances in self-supervised learning have attracted significant attention from both machine learning and neuroscience. This is primarily because self-supervised methods do not require annotated supervisory information, making them applicable to training artificial networks without relying on large amounts of curated data, and potentially offering insights into how the brain adapts to its environment in an unsupervised manner. Although several previous studies have elucidated the correspondence between neural representations in deep convolutional neural networks (DCNNs) and biological systems, the extent to which unsupervised or self-supervised learning can explain the human-like acquisition of categorically structured information remains less explored. In this study, we investigate the correspondence between the internal representations of DCNNs trained using a self-supervised contrastive learning algorithm and human semantics and recognition. To this end, we employ a few-shot learning evaluation procedure, which measures the ability of DCNNs to recognize novel concepts from limited exposure, to examine the inter-categorical structure of the learned representations. Two comparative approaches are used to relate the few-shot learning outcomes to human semantics and recognition, with results suggesting that the representations acquired through contrastive learning are well aligned with human cognition. These findings underscore the potential of self-supervised contrastive learning frameworks to model learning mechanisms similar to those of the human brain, particularly in scenarios where explicit supervision is unavailable, such as in human infants prior to language acquisition.

q-bio.NC↗

arXiv2023

Analysis of Pleasantness Evoked by Various Airborne Ultrasound Tactile Stimuli Using Pairwise Comparisons and the Bradley-Terry Model

The presentation of a moving tactile stimulus to a person's forearm evokes a pleasant sensation. The speed, intensity, and contact area of the strokes should be systematically changed to evaluate the relationship between pleasantness and tactile stimuli in more detail. Studies have examined the relationship between stroking stimulation and pleasant sensations using airborne ultrasound tactile displays. The ultrasound-based method has the advantage of reproducible control of the speed, intensity, and contact area of the stimulus. In this study, we prepared new stimuli focusing on the modulation methods and the contact area and aimed to clarify their relationship with pleasantness in more detail. Evaluating subjective sensations, such as pleasantness, numerically and consistently is challenging, warranting evaluation based on comparison. We propose a stimulus evaluation method that combines rough evaluation using Likert scales, detailed evaluation using pairwise comparisons, and quantification of comparison data using the Bradley--Terry model. As a result, we confirmed that the stimulus using lateral modulation and that with a large contact area used in this study were more pleasant than the conventional stimulus for six out of ten participants.

cs.HC↗

arXiv2022

On the Surrogate Gap between Contrastive and Supervised Losses

Contrastive representation learning encourages data representation to make semantically similar pairs closer than randomly drawn negative samples, which has been successful in various domains such as vision, language, and graphs. Recent theoretical studies have attempted to explain the benefit of the large negative sample size by upper-bounding the downstream classification loss with the contrastive loss. However, the previous surrogate bounds have two drawbacks: they are only legitimate for a limited range of negative sample sizes and prohibitively large even within that range. Due to these drawbacks, there still does not exist a consensus on how negative sample size theoretically correlates with downstream classification performance. Following the simplified setting where positive pairs are drawn from the true distribution (not generated by data augmentation; as supposed in previous studies), this study establishes surrogate upper and lower bounds for the downstream classification loss for all negative sample sizes that best explain the empirical observations on the negative sample size in the earlier studies. Our bounds suggest that the contrastive loss can be viewed as a surrogate objective of the downstream loss and larger negative sample sizes improve downstream classification because the surrogate gap between contrastive and supervised losses decays. We verify that our theory is consistent with experiments on synthetic, vision, and language datasets.

cs.LG↗

arXiv2021

Statistical Mechanical Analysis of Catastrophic Forgetting in Continual Learning with Teacher and Student Networks

When a computational system continuously learns from an ever-changing environment, it rapidly forgets its past experiences. This phenomenon is called catastrophic forgetting. While a line of studies has been proposed with respect to avoiding catastrophic forgetting, most of the methods are based on intuitive insights into the phenomenon, and their performances have been evaluated by numerical experiments using benchmark datasets. Therefore, in this study, we provide the theoretical framework for analyzing catastrophic forgetting by using teacher-student learning. Teacher-student learning is a framework in which we introduce two neural networks: one neural network is a target function in supervised learning, and the other is a learning neural network. To analyze continual learning in the teacher-student framework, we introduce the similarity of the input distribution and the input-output relationship of the target functions as the similarity of tasks. In this theoretical framework, we also provide a qualitative understanding of how a single-layer linear learning neural network forgets tasks. Based on the analysis, we find that the network can avoid catastrophic forgetting when the similarity among input distributions is small and that of the input-output relationship of the target functions is large. The analysis also suggests that a system often exhibits a characteristic phenomenon called overshoot, which means that even if the learning network has once undergone catastrophic forgetting, it is possible that the network may perform reasonably well after further learning of the current task.

stat.ML↗

arXiv2019

Monte Carlo study of the critical properties of noncollinear Heisenberg magnets: $O(3)\times O(2)$ universality class

The critical properties of the antiferromagnetic Heisenberg model on the three-dimensional stacked-triangular lattice are studied by means of a large-scale Monte Carlo simulation in order to get insight into the controversial issue of the criticality of the noncollinear magnets with the $O(3)\times O(2)$ symmetry. The maximum size studied is $384^3$, considerably larger than the sizes studied by the previous numerical works on the model. Availability of such large-size data enables us to examine the detailed critical properties including the effect of corrections to the leading scaling. Strong numerical evidence of the continuous nature of the transition is obtained. Our data indicates the existence of significant corrections to the leading scaling. Careful analysis by taking account of the possible corrections yield critical exponents estimates, $α=0.44(3)$, $β=0.26(2)$, $γ=1.03(5)$, $ν=0.52(1)$, $η=0.02(5)$, and the chirality exponents $β_κ=0.40(3)$ and $γ_κ=0.77(6)$, supporting the existence of the $O(3)$ chiral (or $O(3)\times O(2)$) universality class governed by a new `chiral' fixed point. We also obtain an indication that the underlying fixed point is of the focus-type, characterized by the complex-valued correction-to-scaling exponent, $ω=0.1^{+0.4}_{-0.05} + i\ 0.7^{+0.1}_{-0.4}$. The focus-like nature of the chiral fixed point accompanied by the spiral-like renormalization-group (RG) flow is likely to be the origin of the apparently complicated critical behavior. The results are compared and discussed in conjunction with the results of other numerical simulations, several distinct types of RG calculations including the higher-order perturbative massive and massless RG calculations and the nonperturbative functional RG calculation, and the conformal-bootstrap program.

cond-mat.str-el↗

arXiv2019

A Wrapped Normal Distribution on Hyperbolic Space for Gradient-Based Learning

Hyperbolic space is a geometry that is known to be well-suited for representation learning of data with an underlying hierarchical structure. In this paper, we present a novel hyperbolic distribution called \textit{pseudo-hyperbolic Gaussian}, a Gaussian-like distribution on hyperbolic space whose density can be evaluated analytically and differentiated with respect to the parameters. Our distribution enables the gradient-based learning of the probabilistic models on hyperbolic space that could never have been considered before. Also, we can sample from this hyperbolic probability distribution without resorting to auxiliary means like rejection sampling. As applications of our distribution, we develop a hyperbolic-analog of variational autoencoder and a method of probabilistic word embedding on hyperbolic space. We demonstrate the efficacy of our distribution on various datasets including MNIST, Atari 2600 Breakout, and WordNet.

stat.ML↗

arXiv2017

Concept Formation and Dynamics of Repeated Inference in Deep Generative Models

Deep generative models are reported to be useful in broad applications including image generation. Repeated inference between data space and latent space in these models can denoise cluttered images and improve the quality of inferred results. However, previous studies only qualitatively evaluated image outputs in data space, and the mechanism behind the inference has not been investigated. The purpose of the current study is to numerically analyze changes in activity patterns of neurons in the latent space of a deep generative model called a "variational auto-encoder" (VAE). What kinds of inference dynamics the VAE demonstrates when noise is added to the input data are identified. The VAE embeds a dataset with clear cluster structures in the latent space and the center of each cluster of multiple correlated data points (memories) is referred as the concept. Our study demonstrated that transient dynamics of inference first approaches a concept, and then moves close to a memory. Moreover, the VAE revealed that the inference dynamics approaches a more abstract concept to the extent that the uncertainty of input data increases due to noise. It was demonstrated that by increasing the number of the latent variables, the trend of the inference dynamics to approach a concept can be enhanced, and the generalization ability of the VAE can be improved.

stat.ML↗