arXiv subjects

Nicola Toschi

Publications and source records attributed to Nicola Toschi.

At least 19 recordsLinked to original sources

arXiv2026

Hierarchical Pooling for Sheaf Neural Networks

Sheaf Neural Networks (SNNs) generalize Graph Neural Networks (GNNs) by replacing scalar node signals with stalk-valued signals and by using restriction maps to measure compatibility across edges. Unlike standard graph diffusion, which encourages neighboring node features to become similar, sheaf diffusion promotes consistency through the restriction maps and can therefore model more general relationships between neighboring nodes. However, existing sheaf neural architectures mainly operate at a fixed graph resolution and do not provide a principled pooling mechanism for building hierarchical representations. In this paper, we introduce Hierarchical Sheaf Pool (HiSP), a sheaf-aware pooling framework based on local spectral coarsening. Given a partition of the graph, HiSP constructs each coarse stalk by projecting fine stalk-valued features onto the low-frequency eigenmodes of the cluster-internal sheaf Laplacian. These local modes define a cochain-level prolongation map, which allows the fine sheaf energy to be represented on the coarse space through a Galerkin operator. We further analyze the approximation induced by coarsening by separating truncation loss, due to discarded local modes, from realization loss, due to representing the projected operator as a coarse sheaf. Finally, we implement HiSP as a GNN pooling layer compatible with SNNs and provide a PyG implementation supporting batching, lifted sheaf Laplacians, and hierarchical architectures.

cs.LG↗

arXiv2026

Retrieval-Based Brain Decoding by Alignment, not Complexity

A prominent theory in cognitive science suggests that concepts in the brain are organized as high-dimensional vectors, with semantic meaning captured by directions and relative angles in this space. Brain decoding is the effort of reconstructing or retrieving stimuli (or their representations) from neural activity and involves finding a function that approximates how the brain represents concepts. This motivates the investigation of contrastive objectives as biologically plausible candidates to reverse the brain loss function. In this work, we study how functional MRI (fMRI) activity can generally be mapped with the embedding spaces of foundation models in vision, language, and audio. Although neural computations are highly non-linear at the microscale, fMRI measurements average signals across space and time, further smoothed by noise, effectively linearizing the observable representation. Consistent with these views, our experiments across multiple datasets demonstrate that linear contrastive decoders consistently outperform ridge regression and standard non-linear alternatives, and that these results generalize across images, text, and sound. These findings indicate that decoding gains arise more from the choice of training objective than from architectural complexity, pointing to contrastive-linear models as a principled strategy for brain decoding.

q-bio.NC↗

arXiv2026

Mapping Whisper Representations to Human ECoG Responses with Interpretable Time-Resolved Neural Encoding

Understanding how speech foundation models relate to human cortical activity is a key challenge for computational neuroscience. Here, we investigate how internal representations from Whisper predict intracranial ECoG responses during naturalistic speech perception. We introduce a time-resolved neural encoder that combines speech embeddings with a recurrent temporal model and soft attention, allowing us to examine layer-wise brain alignment. Intermediate Whisper layers provide the strongest correspondence with neural activity, supporting a hierarchical match between model representations and cortical speech processing. Comparisons with baselines show that high-resolution ECoG responses benefit from temporally structured modelling beyond linear mappings from the same speech representations. In addition, attention maps reveal temporally local alignment between speech embeddings and neural responses, while a phonemic interpretability analysis identifies anatomically coherent phoneme-category organization among encoding-informative electrodes. Together, these results suggest that speech foundation models offer a useful framework for studying time-resolved cortical speech representations.

q-bio.NC↗

arXiv2026

Hyperbolic Graph Neural Networks Under the Microscope: The Role of Geometry-Task Alignment

Many complex networks exhibit hierarchical, tree-like structures, making hyperbolic space a natural candidate wherein to learn representations of them. Based on this observation, Hyperbolic Graph Neural Networks (HGNNs) have been widely adopted as a principled choice for representation learning on tree-like graphs. In this work, we question this paradigm by proposing the additional condition of geometry--task alignment, i.e., whether the metric structure of the target follows that of the input graph. We theoretically and empirically demonstrate the capability of HGNNs to recover low-distortion representations on regression problems, and show that their geometric inductive bias becomes helpful when the problem requires preserving metric structure. By jointly analyzing predictive performance and embedding distortion, we further show that HGNNs gain an advantage on link prediction, a naturally geometry-aligned task, whereas this advantage largely disappears on standard node classification benchmarks, which are typically not geometry--aligned. Overall, our findings shift the focus from only asking "Is the graph hyperbolic?" to also questioning "Is the task aligned with hyperbolic geometry?", showing that HGNNs consistently outperform Euclidean models under such alignment, while their advantage vanishes otherwise.

cs.LG↗

arXiv2026

Seeing the imagined: a latent functional alignment in visual imagery decoding from fMRI data

Recent progress in visual brain decoding from fMRI has been enabled by large-scale datasets such as the Natural Scenes Dataset (NSD) and powerful diffusion-based generative models. While current pipelines are primarily optimized for perception, their performance under mental-imagery remains less well understood. In this work, we study how a state-of-the-art (SOTA) perception decoder (DynaDiff) can be adapted to reconstruct imagined content from the Imagery-NSD benchmark. We propose a latent functional alignment approach that maps imagery-evoked activity into the pretrained model's conditioning space, while keeping the remaining components frozen. To mitigate the limited amount of matched imagery-perception supervision, we further introduce a retrieval-based augmentation strategy that selects semantically related NSD perception trials. Across four subjects, latent functional alignment consistently improves high-level semantic reconstruction metrics relative to the frozen pretrained baseline and a voxel-space ridge alignment baseline, and enables above-chance decoding from multiple cortical regions. These results suggest that semantic structure learned from perception can be leveraged to stabilize and improve visual imagery decoding under out-of-distribution conditions.

q-bio.NC↗

arXiv2026

Simple Models, Rich Representations: Visual Decoding from Primate Intracortical Neural Signals

Understanding how neural activity gives rise to perception is a central challenge in neuroscience. We address the problem of decoding visual information from high-density intracortical recordings in primates, using the THINGS Ventral Stream Spiking Dataset. We systematically evaluate the effects of model architecture, training objectives, and data scaling on decoding performance. Results show that decoding accuracy is mainly driven by modeling temporal dynamics in neural signals, rather than architectural complexity. A simple model combining temporal attention with a shallow MLP achieves up to 70% top-1 image retrieval accuracy, outperforming linear baselines as well as recurrent and convolutional approaches. Scaling analyses reveal predictable diminishing returns with increasing input dimensionality and dataset size. Building on these findings, we design a modular generative decoding pipeline that combines low-resolution latent reconstruction with semantically conditioned diffusion, generating plausible images from 200 ms of brain activity. This framework provides principles for brain-computer interfaces and semantic neural decoding.

q-bio.NC↗

arXiv2025

Bio-integrated $μ$Bots with Overtone Ultra-Wideband Magnetoelectric Antennas for Wireless Telemetry

Implantable and wearable devices require antennas that are both miniaturized and efficient, yet conventional designs are constrained by narrow bandwidth and orientation sensitivity. We report overtone ultra-wideband magnetoelectric (OUWB-ME) antennas that exploit higher order acoustic modes in polished silicon substrates to achieve a 22.6 GHz bandwidth in the 3-4 GHz range. Packaged into "$μ$Bots," these magnetoelectric heterostructures bonded with silver nanoparticle inks maintain stable operation under biological loading. In vitro assays confirm the biocompatibility of AlN and the protective role of parylene encapsulation for FeGa. Ex vivo rat and human tissues reshape transmission spectra, identifying reproducible frequency windows near 3.3 and 3.9 GHz. $μ$Bots enable real-time audiovisual telemetry using software-defined radios and exhibit compatibility with 7T MRI. By combining wideband response, robustness to misalignment, and biocompatible packaging, OUWB-ME $μ$Bots provide a scalable platform for wireless bio-integrated communication and telemetry.

physics.app-ph↗

arXiv2025

A Differentiable Model for Optimizing the Genetic Drivers of Synaptogenesis

There is growing consensus among neuroscientists that neural circuits critical for survival are the result of genomic decompression processes. We introduce SynaptoGen, a novel computational framework--member of the Connectome Models family--bringing synthetic biological intelligence closer, facilitating neural biological agent development through precise genetic control of synaptogenesis. SynaptoGen is the first model of its kind offering mechanistic explanation of synaptic multiplicity based on genetic expression and protein interaction probabilities. The framework connects genetic factors through a differentiable function, working as a neural network where synaptic weights equal average numbers of synapses between neurons, multiplied by conductance, derived from genetic profiles. Differentiability enables gradient-based optimization, allowing generation of genetic expression patterns producing pre-wired biological agents for specific tasks. Validation in simulated synaptogenesis scenarios shows agents successfully solving four reinforcement learning benchmarks, consistently surpassing control baselines. Despite gaps in biological realism requiring mitigation, this framework has potential to accelerate synthetic biological intelligence research.

cs.NE↗

arXiv2025

Training Neural Networks by Optimizing Neuron Positions

The high computational complexity and increasing parameter counts of deep neural networks pose significant challenges for deployment in resource-constrained environments, such as edge devices or real-time systems. To address this, we propose a parameter-efficient neural architecture where neurons are embedded in Euclidean space. During training, their positions are optimized and synaptic weights are determined as the inverse of the spatial distance between connected neurons. These distance-dependent wiring rules replace traditional learnable weight matrices and significantly reduce the number of parameters while introducing a biologically inspired inductive bias: connection strength decreases with spatial distance, reflecting the brain's embedding in three-dimensional space where connections tend to minimize wiring length. We validate this approach for both multi-layer perceptrons and spiking neural networks. Through a series of experiments, we demonstrate that these spatially embedded neural networks achieve a performance competitive with conventional architectures on the MNIST dataset. Additionally, the models maintain performance even at pruning rates exceeding 80% sparsity, outperforming traditional networks with the same number of parameters under similar conditions. Finally, the spatial embedding framework offers an intuitive visualization of the network structure.

cs.LG↗

arXiv2025

Transforming Multimodal Models into Action Models for Radiotherapy

Radiotherapy is a crucial cancer treatment that demands precise planning to balance tumor eradication and preservation of healthy tissue. Traditional treatment planning (TP) is iterative, time-consuming, and reliant on human expertise, which can potentially introduce variability and inefficiency. We propose a novel framework to transform a large multimodal foundation model (MLM) into an action model for TP using a few-shot reinforcement learning (RL) approach. Our method leverages the MLM's extensive pre-existing knowledge of physics, radiation, and anatomy, enhancing it through a few-shot learning process. This allows the model to iteratively improve treatment plans using a Monte Carlo simulator. Our results demonstrate that this method outperforms conventional RL-based approaches in both quality and efficiency, achieving higher reward scores and more optimal dose distributions in simulations on prostate cancer data. This proof-of-concept suggests a promising direction for integrating advanced AI models into clinical workflows, potentially enhancing the speed, quality, and standardization of radiotherapy treatment planning.

cs.LG↗

arXiv2024

Bridging Auditory Perception and Language Comprehension through MEG-Driven Encoding Models

Understanding the neural mechanisms behind auditory and linguistic processing is key to advancing cognitive neuroscience. In this study, we use Magnetoencephalography (MEG) data to analyze brain responses to spoken language stimuli. We develop two distinct encoding models: an audio-to-MEG encoder, which uses time-frequency decompositions (TFD) and wav2vec2 latent space representations, and a text-to-MEG encoder, which leverages CLIP and GPT-2 embeddings. Both models successfully predict neural activity, demonstrating significant correlations between estimated and observed MEG signals. However, the text-to-MEG model outperforms the audio-based model, achieving higher Pearson Correlation (PC) score. Spatially, we identify that auditory-based embeddings (TFD and wav2vec2) predominantly activate lateral temporal regions, which are responsible for primary auditory processing and the integration of auditory signals. In contrast, textual embeddings (CLIP and GPT-2) primarily engage the frontal cortex, particularly Broca's area, which is associated with higher-order language processing, including semantic integration and language production, especially in the 8-30 Hz frequency range. The strong involvement of these regions suggests that auditory stimuli are processed through more direct sensory pathways, while linguistic information is encoded via networks that integrate meaning and cognitive control. Our results reveal distinct neural pathways for auditory and linguistic information processing, with higher encoding accuracy for text representations in the frontal regions. These insights refine our understanding of the brain's functional architecture in processing auditory and textual information, offering quantitative advancements in the modelling of neural responses to complex language stimuli.

q-bio.NC↗

arXiv2024

Towards Neural Foundation Models for Vision: Aligning EEG, MEG, and fMRI Representations for Decoding, Encoding, and Modality Conversion

This paper presents a novel approach towards creating a foundational model for aligning neural data and visual stimuli across multimodal representationsof brain activity by leveraging contrastive learning. We used electroencephalography (EEG), magnetoencephalography (MEG), and functional magnetic resonance imaging (fMRI) data. Our framework's capabilities are demonstrated through three key experiments: decoding visual information from neural data, encoding images into neural representations, and converting between neural modalities. The results highlight the model's ability to accurately capture semantic information across different brain imaging techniques, illustrating its potential in decoding, encoding, and modality conversion tasks.

cs.CV↗

arXiv2024

Genetic Motifs as a Blueprint for Mismatch-Tolerant Neuromorphic Computing

Mixed-signal implementations of SNNs offer a promising solution to edge computing applications that require low-power and compact embedded processing systems. However, device mismatch in the analog circuits of these neuromorphic processors poses a significant challenge to the deployment of robust processing in these systems. Here we introduce a novel architectural solution inspired by biological development to address this issue. Specifically we propose to implement architectures that incorporate network motifs found in developed brains through a differentiable re-parameterization of weight matrices based on gene expression patterns and genetic rules. Thanks to the gradient descent optimization compatibility of the method proposed, we can apply the robustness of biological neural development to neuromorphic computing. To validate this approach we benchmark it using the Yin-Yang classification dataset, and compare its performance with that of standard multilayer perceptrons trained with state-of-the-art hardware-aware training method. Our results demonstrate that the proposed method mitigates mismatch-induced noise without requiring precise device mismatch measurements, effectively outperforming alternative hardware-aware techniques proposed in the literature, and providing a more general solution for improving the robustness of SNNs in neuromorphic hardware.

cs.NE↗

arXiv2024

R&B -- Rhythm and Brain: Cross-subject Decoding of Music from Human Brain Activity

Music is a universal phenomenon that profoundly influences human experiences across cultures. This study investigates whether music can be decoded from human brain activity measured with functional MRI (fMRI) during its perception. Leveraging recent advancements in extensive datasets and pre-trained computational models, we construct mappings between neural data and latent representations of musical stimuli. Our approach integrates functional and anatomical alignment techniques to facilitate cross-subject decoding, addressing the challenges posed by the low temporal resolution and signal-to-noise ratio (SNR) in fMRI data. Starting from the GTZan fMRI dataset, where five participants listened to 540 musical stimuli from 10 different genres while their brain activity was recorded, we used the CLAP (Contrastive Language-Audio Pretraining) model to extract latent representations of the musical stimuli and developed voxel-wise encoding models to identify brain regions responsive to these stimuli. By applying a threshold to the association between predicted and actual brain activity, we identified specific regions of interest (ROIs) which can be interpreted as key players in music processing. Our decoding pipeline, primarily retrieval-based, employs a linear map to project brain activity to the corresponding CLAP features. This enables us to predict and retrieve the musical stimuli most similar to those that originated the fMRI data. Our results demonstrate state-of-the-art identification accuracy, with our methods significantly outperforming existing approaches. Our findings suggest that neural-based music retrieval systems could enable personalized recommendations and therapeutic applications. Future work could use higher temporal resolution neuroimaging and generative models to improve decoding accuracy and explore the neural underpinnings of music perception and emotion.

q-bio.NC↗

arXiv2024

Multimodal and multicontrast image fusion via deep generative models

Recently, it has become progressively more evident that classic diagnostic labels are unable to reliably describe the complexity and variability of several clinical phenotypes. This is particularly true for a broad range of neuropsychiatric illnesses (e.g., depression, anxiety disorders, behavioral phenotypes). Patient heterogeneity can be better described by grouping individuals into novel categories based on empirically derived sections of intersecting continua that span across and beyond traditional categorical borders. In this context, neuroimaging data carry a wealth of spatiotemporally resolved information about each patient's brain. However, they are usually heavily collapsed a priori through procedures which are not learned as part of model training, and consequently not optimized for the downstream prediction task. This is because every individual participant usually comes with multiple whole-brain 3D imaging modalities often accompanied by a deep genotypic and phenotypic characterization, hence posing formidable computational challenges. In this paper we design a deep learning architecture based on generative models rooted in a modular approach and separable convolutional blocks to a) fuse multiple 3D neuroimaging modalities on a voxel-wise level, b) convert them into informative latent embeddings through heavy dimensionality reduction, c) maintain good generalizability and minimal information loss. As proof of concept, we test our architecture on the well characterized Human Connectome Project database demonstrating that our latent embeddings can be clustered into easily separable subject strata which, in turn, map to different phenotypical information which was not included in the embedding creation process. This may be of aid in predicting disease evolution as well as drug response, hence supporting mechanistic disease understanding and empowering clinical trials.

cs.LG↗

arXiv2023

4Ward: a Relayering Strategy for Efficient Training of Arbitrarily Complex Directed Acyclic Graphs

Thanks to their ease of implementation, multilayer perceptrons (MLPs) have become ubiquitous in deep learning applications. The graph underlying an MLP is indeed multipartite, i.e. each layer of neurons only connects to neurons belonging to the adjacent layer. In contrast, in vivo brain connectomes at the level of individual synapses suggest that biological neuronal networks are characterized by scale-free degree distributions or exponentially truncated power law strength distributions, hinting at potentially novel avenues for the exploitation of evolution-derived neuronal networks. In this paper, we present ``4Ward'', a method and Python library capable of generating flexible and efficient neural networks (NNs) from arbitrarily complex directed acyclic graphs. 4Ward is inspired by layering algorithms drawn from the graph drawing discipline to implement efficient forward passes, and provides significant time gains in computational experiments with various Erdős-Rényi graphs. 4Ward not only overcomes the sequential nature of the learning matrix method, by parallelizing the computation of activations, but also addresses the scalability issues encountered in the current state-of-the-art and provides the designer with freedom to customize weight initialization and activation functions. Our algorithm can be of aid for any investigator seeking to exploit complex topologies in a NN design framework at the microscale.

cs.NE↗

arXiv2023

Beyond Multilayer Perceptrons: Investigating Complex Topologies in Neural Networks

In this study, we explore the impact of network topology on the approximation capabilities of artificial neural networks (ANNs), with a particular focus on complex topologies. We propose a novel methodology for constructing complex ANNs based on various topologies, including Barabási-Albert, Erdős-Rényi, Watts-Strogatz, and multilayer perceptrons (MLPs). The constructed networks are evaluated on synthetic datasets generated from manifold learning generators, with varying levels of task difficulty and noise, and on real-world datasets from the UCI suite. Our findings reveal that complex topologies lead to superior performance in high-difficulty regimes compared to traditional MLPs. This performance advantage is attributed to the ability of complex networks to exploit the compositionality of the underlying target function. However, this benefit comes at the cost of increased forward-pass computation time and reduced robustness to graph damage. Additionally, we investigate the relationship between various topological attributes and model performance. Our analysis shows that no single attribute can account for the observed performance differences, suggesting that the influence of network topology on approximation capabilities may be more intricate than a simple correlation with individual topological attributes. Our study sheds light on the potential of complex topologies for enhancing the performance of ANNs and provides a foundation for future research exploring the interplay between multiple topological attributes and their impact on model performance.

cs.NE↗

arXiv2023

Decoding visual brain representations from electroencephalography through Knowledge Distillation and latent diffusion models

Decoding visual representations from human brain activity has emerged as a thriving research domain, particularly in the context of brain-computer interfaces. Our study presents an innovative method that employs to classify and reconstruct images from the ImageNet dataset using electroencephalography (EEG) data from subjects that had viewed the images themselves (i.e. "brain decoding"). We analyzed EEG recordings from 6 participants, each exposed to 50 images spanning 40 unique semantic categories. These EEG readings were converted into spectrograms, which were then used to train a convolutional neural network (CNN), integrated with a knowledge distillation procedure based on a pre-trained Contrastive Language-Image Pre-Training (CLIP)-based image classification teacher network. This strategy allowed our model to attain a top-5 accuracy of 80%, significantly outperforming a standard CNN and various RNN-based benchmarks. Additionally, we incorporated an image reconstruction mechanism based on pre-trained latent diffusion models, which allowed us to generate an estimate of the images which had elicited EEG activity. Therefore, our architecture not only decodes images from neural activity but also offers a credible image reconstruction from EEG only, paving the way for e.g. swift, individualized feedback experiments. Our research represents a significant step forward in connecting neural signals with visual cognition.

eess.SP↗