Frontiers in Computational Neuroscience

The -Bonacci weighted sequence space {ell}(,) is introduced for [&ge;] 2, 1 [&le;] < {infty}, and > 0). It is proved to be a Banach space which is isometrically isomorphic to {ell}, with an unconditional Schauder basis. The space is also reflexive and uniformly convex. Its strict inclusion chain, continuous dual, and asymptotic truncation estimate of order - / are also established. Based on this framework, Scaled &#119898-BonacciNet is proposed as a lightweight CNN that rescales Fibonacci-Net channels using the dominant &#119898-Bonacci root and employs depth wise separable convolutions. Tribonacci-Net achieved 96.09% accuracy and a 0.9581 macro F1-score on the 15-class Brain MRI dataset using 0.85 million parameters and 0.366 GFLOPs. Pentanacci-Net attained 95.87% accuracy on the 44-class dataset while reducing parameters and FLOPs by 47.10% and 44.48%, respectively. On the CKD ultrasound dataset, Hexanacci-Net achieved 83.75% accuracy and a 0.7941 macro F1-score, with corresponding reductions of 50.37% and 45.10%. These findings support dominant-root scaling as a structured approach to lightweight CNN design.

Decision-making requires accumulation of evidence over time, yet exactly how neural circuits implement this computation remains debated. Classical models posit that evidence accumulation is realized by ramping neural activity, whereas recent experimental findings suggest an alternative mechanism in which neural activity remains largely silent before abruptly transitioning into an active decision state. This mechanism leaves a fundamental question unanswered: how is evidence accumulated during the neural activity-silent period? To address this question, we trained excitatory-inhibitory recurrent neural networks (RNNs) with short-term synaptic plasticity (STP) on a reaction-time decision task. We found that after training, the networks learned to leverage STP to make decisions in an activity-silent manner. Choice-tuning analysis revealed a clear functional organization of the network, and connection-pruning analysis uncovered the coupling between choice-selective excitatory and inhibitory neural groups. Guided by the architecture of trained RNNs, we constructed a minimal neural circuit model, in which evidence accumulates in facilitated synaptic efficacy that progressively reshapes the circuit's phase portrait, eventually driving the network state across the stability boundary and triggering a rapid transition to a choice state. We further compared activity before a choice between the theoretical STP and Static models to examine the potential energy benefit of activity-silent decision-making. Together, our results suggest STP as a mechanism for evidence accumulation in activity-silent decision-making and offer insights into the design of energy-efficient artificial neural networks.

We present a network model that enables a navigating agent to shape the spatial and temporal structure of its paths through its environment. The model is inspired by the insect fan-shaped body. In it, a recurrent neural network receives an input that encodes an agent's velocity vector. Meanwhile, network activity determines the setpoint of a control loop that influences this velocity vector through motor outputs. The network thus guides future movement using memory of past movement. We illustrate how the network's dynamics determine the trajectory features it extracts to produce the setpoint. We then relate the joint dynamics of network activity and the agent's velocity vector to the structure of the resulting paths. This network can produce paths suitable for different forms of exploration. It can also stabilize movements initiated by lower-level systems. Our analyses illustrate how recurrent neural dynamics and memory can shape behavior when embedded in closed loops.

Astrocytes actively regulate synaptic transmission and neuronal excitability, yet their role in orchestrating macroscopic cortical network regimes and slow-wave oscillations remains an active area of reasearch. This study investigates how bidirectional neuron astrocyte interactions shape emergent population dynamics using a computational network model of excitatory and inhibitory neurons coupled to an astrocyte. The results identify astrocytic feedback topology, rather than astrocytic coupling strength alone, as a key determinant of emergent cortical network dynamics. By systematically dissecting pathway-specific connectivity, it has been shown that the neuronal population driving astrocytic activation and the neuronal population receiving gliotransmission jointly determine whether the network occupies asynchronous irregular (AI), synchronous irregular (SI), synchronous regular(SR), asynchronous regular(AR) or quiescent regimes.Directing gliotransmission selectively onto excitatory neurons consistently promotes population synchrony regardless of the population influencing astrocytic dynamics, whereas selective modulation of inhibitory interneurons induces network quiescence via strong suppression. Under dual-target gliotransmission, network synchrony is dictated by the population driving astrocytic dynamics: excitatory-only drive promotes synchrony, while combined or inhibitory-specific drive preserves asynchronous states. Furthermore, the model reveals that astrocytic signaling kinetics provide an additional temporal control mechanism that regulates the frequency and persistence of self sustained up states.

Electrical stimulation of brain activity has emerged as a key therapeutic intervention for neurological diseases. The success of these approaches relies on precisely tuned stimulation parameters. A major challenge is that these parameters should ideally be adjusted in real-time based on changes in brain activity. Closed-loop controllers address this challenge by monitoring brain activity and adapting stimulation according to ongoing changes in neural dynamics. While machine learning has been helpful in this context, closed-loop controllers typically do not adapt in real-time to fluctuations in brain activity. Here, we propose a novel framework based on reservoir computing to train a closed-loop, real-time neural controller. This approach, termed CLEAR-FORCE, trains the output weights of a recurrent network based on a combination of target signals and feedback from brain activity. Two series of simulations were performed, where the controller was successful at (1) inducing neural oscillations in a population of spiking neurons and (2) suppressing pathological oscillations. This framework provides a promising avenue for the implementation of neural controllers in low-powered, real-time, and adaptive hardware that harnesses machine learning for the personalized treatment of brain disorders.

Competitive spiking networks can form sensory representations through local synaptic plasticity and lateral inhibition. A canonical example is the Diehl-Cook architecture, in which Poisson-encoded sensory inputs drive excitatory neurons through spike-timing-dependent plasticity (STDP), while a separate inhibitory population imposes competition between excitatory units. In most implementations this competitive circuit is fixed, although inhibitory synapses are plastic in cortical circuits and contribute to excitation-inhibition balance. It therefore remains unclear under which conditions biologically motivated inhibitory plasticity can replace fixed lateral competition without disrupting representation learning. Here we study this question in a competitive STDP network trained on MNIST. We compare fixed inhibition with a Vogels-style inhibitory STDP rule, slow homeostatic inhibition, and a budget-constrained slow homeostatic rule that conserves the total inhibitory input received by each excitatory neuron. A fixed-inhibition operating-point map showed that stability depends on the dynamical regime: 56% of tested fixed-inhibition runs completed training, whereas stronger inhibition at higher input gain produced runaway activity. The reference regime g = 2.0, wIE = 10.0, selected for subsequent comparisons, combined complete stability across replicate initializations with high classification performance (84.3 {+/-} 0.6% after 30,000 training examples). In this regime, the budget-constrained rule was stable in all replicate runs when its inhibitory budget matched the fixed-inhibition reference (BI = 3990), reaching 83.5 {+/-} 0.6% accuracy with firing-rate and inhibitory-conductance statistics close to fixed inhibition, whereas a mismatched budget (BI = 7980) produced runaway activity in every replicate. Parameter maps showed that Vogels-derived iSTDP was broadly but not uniformly stable (90% runs completed), while slow homeostatic inhibition was even more broadly stable (96%). Budget-constrained stability depended strongly on the conserved budget, i.e., low and matched budgets were stable or nearly stable, intermediate budgets were mixed, and BI = 7980 failed in all tested runs. These results show that inhibitory plasticity in competitive STDP networks is not governed by a local homeostatic update alone. Stable plastic competition depends jointly on the operating regime and on an inhibitory budget that fixes the total strength of competition while allowing local inhibitory weights to adapt. Mechanistically, budget matching separates local redistribution of inhibitory weights from changes in total inhibitory drive, preventing plasticity from shifting the network away from a stable competitive state. This identifies inhibitory budget matching as a practical design principle for replacing fixed inhibition with plastic, homeostatically regulated competition in STDP-based spiking networks.

Parkinson's disease (PD) gait is characterized by reduced walking speed, short shuffling steps, increased double stance time, and increased step time variability. The most debilitating symptom in PD is freezing of gait (FOG), a context-dependent locomotion arrest, mostly occurring while encountering turns, obstacles, and narrow pathways. FOG is more pronounced during turning, a complex task that challenges motor planning and execution. Here, we simulate this phenomenon using a computational model inspired by the cortico-basal ganglia circuitry, based on reinforcement learning (RL) principles. In the model, the agent was trained to learn a value profile while navigating a path that involved turning. To simulate PD conditions, we clamped the temporal difference (TD) error, typically associated with dopaminergic signaling. Our model results show that compared to healthy controls (HC), PD non-freezers and PD freezers exhibit a reduction in turning velocity and step length, alongside an attenuated compensatory step-width widening at the sharpest turn, more pronounced in PD freezers, who additionally show increased step time variability. These kinematic impairments worsen during tighter turns. Ultimately, this model implicates FOG as a problem of impaired estimation of value over space and could potentially represent a generalized deficit in gait impairments in PD.

Emerging research on desynchronization reveals strong relations to memory formation and recollection with much of the evidence originating from studies of the hippocampus, a brain region central to memory processing. However, growing evidence emphasizing the role of desynchronization is accompanied by challenges, most notably the synchronization/desynchronization conundrum. The conundrum arises from an apparent contradiction between learning and Shannon's information theory: spike-timing-dependent plasticity (STDP) acquires information via synchronized populational activity whereas information theory suggests that information channel is encoded through higher variability. Both frameworks are well-founded. Learning is widely accepted as the mathematical abstraction of STDP while Shannon's information theory showcases its power ubiquitously across fields from communication systems to neuroscience. Here, we propose that learning and Shannon's information theory are not necessarily opposing principles. Rather, learning plays a critical role in the process of synchronization and desynchronization. To justify our claim, we conduct computational simulations based on a Hodgkin-Huxley neuronal model coupled with NMDA, AMPA, and GABA synapses with inputs of biologically realistic spiking data. Our results show that enabling STDP significantly influences synchronization and desynchronization dynamics. Specifically, we find that STDP and desynchronization form a regulatory loop, in which STDP regulates the level of synchrony, and synchrony regulates subsequent learning strength. Further study reveals STDP is capable of switching neurons from synchronization to desynchronization. This framework mitigates the synchronization/desynchronization conundrum by clarifying the relationship between desynchronization and learning, and offers a new perspective on how synchrony contributes to memory formation and recollection.

Online identification of nonlinear dynamical systems faces two challenges. Noisy, partial observations make the latent state difficult to estimate. Local attractors can also keep passive trajectories in regions that reveal little about the unknown dynamics. Both challenges commonly arise in closed-loop neural experiments where stimulation must be chosen from noisy spike counts, while neural activity may remain near a stable state unless stimulation drives it elsewhere. We address this problem by actively selecting inputs to collect informative observations during the experiment. We introduce PALDI (Planning-based Active Latent Dynamics Identification), which combines extended-Kalman state prediction, single-step Laplace state updates, blockwise Gaussian parameter updates, and finite-horizon planning. Its parameter expected-information-gain surrogate combines dynamics sensitivity with state Fisher information and discounts apparent parameter information when latent states are uncertain. We compare PALDI with nonadaptive inputs, a myopic policy, and two active baselines: the one-step D-optimal method FLEX and receding-horizon curiosity with uncertainty sampling. On a bistable Wilson-Cowan model of excitatory and inhibitory populations, PALDI approaches the true-model reference and remains strongest as the observation signal-to-noise ratio decreases or the readout geometry becomes biased. We then test identification in a spiking decision network where we use a reduced dynamics and observation model to approximate the data-generating process. PALDI reaches a median held-out prediction R2 of 0.90, close to a fitted reference, and its model gives the highest change-of-mind success among identification policies. Information-guided finite-horizon planning is a practical strategy for identifying neural circuit dynamics with closed-loop stimulation.

Alzheimers Disease (AD) is characterised by the progressive depletion of Acetylcholine (ACh), a neurotransmitter essential for cognition and synaptic communication. Docosahexaenoic acid (DHA), the predominant omega 3 polyunsaturated fatty acid in neuronal membranes, raises cerebral ACh and supports cholinergic transmission, but its therapeutic efficacy is limited by poor blood brain barrier permeability and slow passive diffusion. This work addresses these limitations through a closed form mathematical model coupling the spatiotemporal dynamics of DHA diffusion and ACh synthesis under an externally applied static magnetic field. Because DHA is diamagnetically anisotropic, it aligns along the field axis through diamagnetic torque while body cells form field aligned linear arrays; together these effects raise the effective diffusion coefficient of DHA. At a representative observation point, 0.5cm, magnetic field assisted transport reduces the time for ACh to reach a therapeutic threshold by 83.3%, from 10.55 s to 1.76 s. This reduction is spatially uniform across the modelled range of 0 to 3.0 cm from the source, with the absolute saving growing from 6.71 s at the source to 19.21 s at 3.0 cm. Co-evolution analysis reveals the mechanism: by the end of a 2.0 s window the field elevates DHA 7.39 fold relative to baseline, driving ACh to 133.7% of threshold versus only 18.1% without the field. A Brain Computer Interface (BCI)-based closed loop dosing framework is additionally proposed for future clinical validation. These findings establish a quantitatively grounded theoretical basis for magnetically enhanced DHA therapy as an approach to cholinergic restoration in AD.

Face recognition ability varies enormously across humans. Individuals with face-blindness (prosopagnosia) struggle to recognize even close family members while super-recognizers can identify strangers with exceptional accuracy. Where do ANNs fall within the human face recognition spectrum? Here, we administered the standardized tests used to characterize human face recognition ability to a diverse set of models, allowing us to contextualize a model's performance within the distribution of human behavior. We found that the majority 55% of models to be classified as face-blind and that even the best face-trained models do not cross the human threshold to be considered a super-recognizer. Interrogating the internal representations of these models showed that models that performed well on standardized face recognition tasks were more identity-selective and viewpoint-invariant. We also found that low-performing models did contain some identity information in independent representational subspaces. Removing viewpoint-dependent subspace improved face recognition abilities in 49 of the 53 models tested. Targeted unit ablations further identified opposing contributions, with viewpoint-dependent units disrupting identity coding and viewpoint-tolerant units supporting it. Together, our results show that most AI models are face-blind with worse face recognition ability than humans, and demonstrate how differences across AI models can be used generate testable hypotheses about the computational basis of human face recognition in humans which can then be probed in future studies.

How does the brain generate sequences of discrete elements, such as actions or words, that follow an abstract rule (a grammar)? Motivated by an experiment in which monkeys learned to draw multi-object images according to grammatical rules, we trained a recurrent neural network to generate sequences Am Bn: producing shape A m times, then B n times, or vice versa. Our findings are threefold. First, the network plans the sequence by moving to an initial state encoding m and n along orthogonal directions for upcoming execution. Second, the network repeats each shape by progressing along a shape-specific manifold. Third, transitions from Am to Bn follow parallel trajectories between manifolds across values of n, thus preserving n. Each of these is an experimentally testable prediction. The network generalizes to longer novel sequences by starting further along the same orthogonal directions. These results reveal a neural circuit implementation of a symbolic algorithm for generating rule-based action sequences.

The dentate gyrus' principal neuron, the granule cell (GC), plays a key role in pattern separation, transforming similar inputs into dissimilar outputs. We used a new measure based on information theory to evaluate this transformation in an experimentally derived in silico GC model with realistic ion channel distributions across its dendrites. We found that model performance aligned well with current-clamp recordings from mouse brain slices. However, it was also possible to find artificial parameter combinations in the model that functionally outperformed the recorded cells while matching their action potential shapes. Rather than maximising computational power, the biologically realistic GC model balanced energy efficiency against increased functionality in a Pareto optimal way: No possible parameter combination was found to improve one of the objectives without impairing the other. Interestingly, immature GCs adopted a different weighted trade-off with a stronger emphasis on energy efficiency than pattern separation, but they remained Pareto optimal over their putative developmental trajectory. Our results support the idea that evolution favours neurons that can satisfy functional and energetic demands simultaneously.

Cephalopod skin exhibits a wide range of dynamic visual displays, including traveling waves, light/dark flashes, and flickering, all generated by muscle-regulated expansion and retraction of pigment-filled elastic sacs known as chromatophores. While central neural control is essential for high-level body-wide pattern coordination, experimental evidence from excised and denervated skin reveals that intricate spatiotemporal patterns can also emerge spontaneously in the skin. This raises questions about the respective roles of central versus peripheral control. To determine the simplest control necessary to generate such patterns, we developed ChromatoNet, a biophysical model of chromatophore arrays that includes mechanical interactions on 1- and 2-dimensional lattices with the excitability properties of radial muscles. With minimal external input, our modified Morris--Lecar model reproduces a rich repertoire of skin dynamics, including spontaneous flickering as well as traveling and spiral waves. Both spontaneous and stimulus-evoked waves emerge at an intrinsic resonant frequency, determined by model parameters. Propagation speeds remain stable across conditions and are consistent with the range observed in denervated skin preparations. The model generates diverse output states in response to local couplings and the time-dependence and shape of the input stimulation, with little top-down direction. The results provide a proof-of-principle that chromatophore/muscle dynamics with minimal central neural driving can produce rich dynamic patterns.

Biophysically detailed models of human cortical microcircuit provide key insights into brain function and disease biomarkers, yet their computational cost lengthens simulation runtime and limits scalability. To address this, we systematically evaluated recent simplification methods for morphological reduction and synapse merging in terms of preserving spiking and electroencephalographic (EEG) properties of human cortical microcircuit models. At the single- cell level, reduced compartmental models accurately reproduced spiking behaviors. However, at the microcircuit level, while achieving a 5-fold speed-up, morphological reduction failed to preserve spectral dynamics of spiking and EEG despite maintaining average baseline and response firing rates. Conversely, applying synapse merging alone on the full morphology successfully reproduced both spiking activity and EEG dynamics, when constrained to [&le;] 50 m spatial intervals, and yielded a 2-fold speed-up. These results were consistent across both cortical layer 2/3 and layer 5 models, and in depression models with reduced somatostatin interneuron inhibition. Our study identifies model simplification methods that speed up cortical microcircuit simulations while preserving the accuracy of the simulated brain signals.

Whole-brain connectomes are increasingly used as wiring diagrams for spiking simulations and as structural priors for artificial networks, usually after an ad hoc synapse-count threshold whose effect on validated function has never been measured. We define function-preserving sparsification - removing edges under a fixed budget while preserving validated behaviour and population responses - and study it in the leaky integrate-and-fire whole-brain model of Shiu et al. (2024). Six operators - weak-synapse pruning, uniform and weight-proportional random removal, degree-preserving rewiring, weight shuffling and a task-aware criterion - were compared at identical edge budgets across three validated circuits (sugar-evoked proboscis extension, its bitter suppression, grooming pathway specificity), silent and spontaneously active network states, and multiple seeds. Weak-synapse pruning preserves 79% of feeding output with 6% of edges, with a cliff at k* of approximately 9 synapses, whereas random, degree-preserving and weight-shuffled controls at the same budget abolish output, and weight-proportional random sampling survives only while it retains the strong-edge set: function resides in the conjunction of wiring placement and strength, not in degree statistics or synaptic mass. The cliff is task-dependent - grooming specificity collapses below k = 2 through disinhibition. In the active state failure changes character: output over-excites instead of vanishing, and bitter suppression is lost progressively from k = 5 and reverses at k = 31. A task-aware criterion preserves its training stimulus at 0.4% of edges but loses held-out inhibition first. These results provide a first benchmark for connectome sparsification and show that what must be kept depends on task and network state.

The results in this article demonstrate that the central nervous system CNS plans trajectories in the vertical plane by minimizing a Generalized Dynamic Jerk GJerk cost function, which we have proposed as an advancement of the traditional Minimum Jerk cost function. While the traditional Jerk cost function generates smooth trajectories by minimizing the third derivative of joint angle over time, it neglects real dynamic constraints most notably the effect of gravity. According to our engineering interpretation, the arm functions as a constrained manipulator where gravity acts as a non-linear, state-dependent bias; thus, the asymmetry in velocity profiles is not an anomaly but an optimal adaptation to the gravitational field. Consequently, it predicts mathematically symmetrical upward and downward movements, even in contexts where gravitational forces play a decisive role. The proposed GJerk cost function extends the classical framework by incorporating position-dependent dynamic coefficients into the cost function, thereby embedding gravitational effects directly within the planning process. As a result, the generated movements are not only smooth in a mathematical sense but also physically and mechanically consistent with the actual torque and energy demands imposed by the musculoskeletal system. The GJerk cost function reduces angular change in high-gravitational-torque regions because it makes the joint configuration with the highest gravitational torque more expensive within the cost function, prompting the planner to choose a slower and more conservative movement. This does not physically dissipate energy, but it acts as a behavioral damping mechanism, similar to how humans naturally slow down when approaching a high load or potentially dangerous configuration. In this way, the GJerk cost function reduces abrupt changes in joint kinetic energy and prevents sudden jumps in acceleration, ultimately producing a smoother and more stable trajectory that aligns with both biomechanical reality and observed human behavior. Within this framework, upward movements become inherently more costly than downward ones, penalizing configurations that require large torques to counteract gravity. This induces a biomechanically meaningful asymmetry, ensuring the resulting trajectories reflect the body's true mechanical constraints rather than mere geometric smoothness. Our empirical results demonstrate that the GJerk cost function provides a more realistic and physiologically grounded account of motor planning in the vertical plane, linking smoothness optimization with genuine dynamic requirements and offering a closer approximation to biological motor behavior

The basal ganglia (BG), thalamus and cortex form the BG-thalamo-cortical (BGTC) network, which is essential for voluntary movement and centrally involved in Parkinson's disease (PD). In both patients and animal models, BG neuronal activity exhibits exaggerated oscillatory synchronization in the beta frequencies (13-30 Hz). Theoretical studies have proposed multiple network mechanisms for the generation of these abnormal beta oscillations. However, key properties, such as frequency and power distribution across BG nuclei, vary substantially among patients and between animal models (rodents vs. non-human primates). This variability complicates direct comparisons between theoretical models and experimental data, and questions whether distinct neuronal mechanisms may underlie beta oscillations across species. In an experimentally constrained BGTC network model with species-specific parameters (synaptic, neuronal, and network properties), we evaluate the features of abnormal beta oscillatory activity generated by different mechanisms. The network's negative feedback loops serve as potential sources of spontaneous oscillations. Using rodent- or primate-constrained models, we compare the spectral properties of oscillatory activity across network populations for each loop and derive the expected phase relationships between populations to align predictions with existing rodent data and propose testable hypotheses for primates. We also demonstrate how oscillation frequency can be modulated when multiple generation mechanisms interact as coupled oscillators in the full network. Our results, combined with observed cross-species beta oscillations characteristics, suggest that abnormal beta oscillations likely arise from distinct mechanisms in rodents and primates.

Large-scale biophysically-detailed models of the brain are instrumental in unraveling how cellular, synaptic, and circuit-level mechanisms contribute to neural dynamics. Such models are built around large populations of single-neuron models whose morphologies must preserve physiological variability to reproduce integrative properties and firing patterns observed experimentally. Experimental reconstruction of morphologies is a time-consuming and resource-intensive process, unlikely to supply them at the required scale. Several data-driven methods for morphology synthesis have been developed, which nonetheless rely on extensive datasets for parameter estimation. To address this limitation, we developed NeuWalk, to our knowledge the first public Python library for morphology synthesis based on branching-and-annihilating, biased-and-correlated random walks. It estimates branching and annihilation rates from Sholl intersections and, optionally, bifurcation counts, while also ensuring their replication in synthetic morphologies; user-specified combinations of biases, implementing somatofugality, self-avoidance, spatial competition, and branching-angle control, sculpt morphological patterns characteristic of each neuron class. We then validated the method on mitral and middle tufted cells of the olfactory bulb, semilunar and pyramidal cells of the anterior piriform cortex, and neocortical pyramidal neurons: means and variances of Sholl intersections, bifurcation counts, and total dendritic length were statistically indistinguishable between synthetic and natural morphologies, with the single exception of its variance in apical dendrites of piriform cortex pyramidal neurons. As Sholl intersections and bifurcation counts are commonly reported in the literature, and biases can be inferred from published images or qualitative descriptions, NeuWalk requires no immediate access to experimental reconstructions, striking a trade-off between data-driven parameterization and empiricism without sacrificing morphological accuracy.

Nonlinear neural networks develop structured internal representations, yet how their geometry is determined by the tasks being learned remains poorly understood. Here, we develop an analytical framework for piecewise-linear feedforward and recurrent networks that links learning, activation-region structure, and neural selectivity. We show that, at any stationary point of gradient-aligned learning, a nonlinear network decomposes into local linear regressions over its activation regions. Deviations from the corresponding least-squares solutions are jointly constrained by network weights shared across regions and vanish in the low-error regime, yielding an approximate piecewise least-squares decomposition of the task. This structure admits a decision-tree interpretation, which we recover numerically by fitting trees to predict network activation patterns from the input. We further derive how the regions in which neurons are active determine the task statistics they capture, thereby organizing neural selectivity into distinct subpopulations. The resulting predictions closely match the selectivity geometry observed in simulated networks and two empirical neural datasets. Finally, we show that neural baseline regulates activation pattern diversity, placing networks along a continuum between coarse, generalizing representations and fine-grained, expressive representations. Together, these results establish activation regions as a unifying framework for describing how nonlinear networks decompose task structure into local computations and how those computations can be revealed through neural activity.