Biology subjects

Albesa-Gonzalez, A.

Publications and source records attributed to Albesa-Gonzalez, A..

4 recordsLinked to original sources

bioRxiv2025

Homeostatic Binary Networks: A simple framework for learning with overlapping patterns

Memories are rarely stored in isolation: experiences overlap in time and context, leading to neuronal activity patterns that share elements across episodes. While such overlap supports generalization and abstraction, it also increases interference and threatens representational stability. Here we introduce Homeostatic Binary Networks (HBNs), a minimal recurrent framework that combines binary activity, adjustable inhibition, Hebbian learning, and homeostatic plasticity to address these challenges. First, we formalize an Episode Generation Protocol (EGP) that creates compositional episodes with controllable overlap and noise, and define a corresponding semantic structure as conditional probabilities between concepts. We then show analytically and through simulations that recurrent synapses converge to conditional firing probabilities, thereby encoding asymmetric semantic relationships across concepts. These recurrent dynamics enable reliable recall and replay of overlapping episodes without representational collapse. Finally, by incorporating feed-forward plasticity with a neuronal maturity mechanism, output neurons form selective receptive fields in a one-shot manner and refine them through replay, yielding robust unsupervised classification of overlapping episodes. Together, our results demonstrate how simple principles such as neural and synaptic competition can support the stable representation and organization of overlapping memories, providing a mechanistic bridge between episodic and semantic structure in memory systems.

neuroscience↗

bioRxiv2025

From episodes to concepts and back: Semantic representations in episodic memory enhance recall, replay, and compositional consolidation

Semantic knowledge is thought to emerge through the consolidation of episodic experience, yet the biological mechanisms by which reusable semantic representations are extracted from complex episodes remain unclear. Conversely, semantic representations can themselves be found within the medial temporal lobe, raising the questions of how they arise there and why structured semantic overlap should benefit an episodic memory system. We propose that replay establishes a shared semantic vocabulary between MTL and CTX through two complementary processes. First, episodic replay enables neocortex to extract reusable semantic concepts from overlapping episodes via a form of dictionary learning. Second, spontaneous cortical activity replays these learned concepts back to MTL, allowing semantic representations to become part of the episodic code. We implement this framework in a circuit model using local Hebbian plasticity and homeostatic competition. The model reproduces key experimental observations, including concept-cell-like representations, increased overlap between related memories following consolidation, and synaptic reorganization associated with systems consolidation. Functionally, embedding semantic representations within episodic memory improves recall by providing an error-correcting scaffold and enhancing replay consistency, inducing a generalization bias in the consolidation of compositional representations. This further explains experimental effects hard to reconcile with standard accounts of consolidation, such as the impact of disrupting semantics in episodic recall and semantic deficits following medial temporal lobe lesions. More broadly, our results suggest that consolidation progressively builds a shared vocabulary between episodic and semantic memory, allowing structure extracted from past experience to organize the representation and consolidation of future memories.

neuroscience↗

bioRxiv2023

Learning with filopodia and spines

Filopodia are thin synaptic protrusions that have been long known to play an important role in early development. It has recently been found that they are more abundant in the adult cortex than previously thought, and more plastic than spines (button-shaped mature synapses). Inspired by these findings, we introduce a new model of synaptic plasticity that jointly describes learning of filopodia and spines. The model assumes that filopodia exhibit additive learning, which is highly competitive and volatile. At the same time, it proposes that if filopodia undergo sufficient potentiation they consolidate into spines, and start following multiplicative learning dynamics. This makes spines more stable and sensitive to the fine structure of input correlations. We show that our learning rule has a selectivity comparable to additive spike-timing-dependent plasticity (STDP) and represents input correlations as well as multiplicative STDP. We also show how it can protect previously formed memories and act as a synaptic consolidation mechanism. Overall, our results provide a mechanistic explanation of how filopodia and spines could cooperate to overcome the difficulties that these separate forms of learning (additive and multiplicative) each have. Author SummaryChanges in the strength of synaptic connections between neurons are the basis of learning in biological and artificial networks. In animals, these changes can only depend on locally available signals, and are usually modeled with learning rules. Based on recent discoveries on filopodia, a special type of synaptic structure, we propose a new learning rule called Filopodium-Spine spike-timing-dependent-plasticity. Our rule proposes that filopodia follow additive STDP and spines (mature synapses) multiplicative STDP. We show that our model overcomes classic difficulties that these learning rules have separately, like the absence of stability or specificity, and can also be seen as a first stage of synaptic consolidation.

neuroscience↗

bioRxiv2022

Theta oscillations optimize information rate in the hippocampus

Low-frequency oscillations shape how neurons sample their synaptic inputs, regulating information exchange across networks. In the hippocampus, theta-band oscillations (3-8 Hz) reorganize cortical input signals temporally, resulting in a phase code. However, the reason hippocampal oscillations are limited to low frequencies like the theta band remains unclear. Here, we derive a theoretical framework for neuronal phase coding to show that realistic noise levels create a trade-off between sampling speed (controlled by oscillation frequency) and encoding precision in hippocampal neurons. This speed-precision trade-off produces a maximum in information rate within the theta band of ~1-2 bits/s. Additionally, we demonstrate that our framework explains other key hippocampal properties, such as the preservation of theta along the dorsoventral axis despite various physiological gradients, and the modulation of theta frequency and amplitude by the animals running speed. Extending our analysis to extra-hippocampal areas, we propose that theta oscillations may also support efficient encoding of stimuli in visual cortex and olfactory bulb. More broadly, we lay the groundwork for rigorously studying how system constraints determine optimal sampling frequency regimes for phase coding neurons in biological and artificial brains. Author SummaryThe rodent hippocampus exhibits prominent oscillations in the theta band (3-8 Hz) during exploration, enabling individual neurons to rhythmically sample and represent sensory signals from the cortex. However, the reason behind the specific frequency of this hippocampal rhythm has remained unclear. In this study, we developed a biologically-based theoretical framework to demonstrate that neurons using oscillations to efficiently sample noisy signals encounter a trade-off between their sampling speed (i.e., oscillation frequency) and their coding precision (i.e., reliability of encoding). Notably, our findings reveal that this trade-off is optimized precisely within the theta band, while also providing insights into other fundamental features. In conclusion, we offer an explanation grounded in efficient coding for why hippocampal oscillations are confined to the theta band and establish a foundation for exploring how the properties of individual neurons determine optimal sampling frequencies in specific neural circuits.

neuroscience↗