Optimal rate-variance coding due to firing threshold adaptation near criticality

Recurrently connected neuron populations play key roles in sensory perception and memory storage across various brain regions. While these populations are often assumed to encode information through firing rates, this method becomes unreliable with weak stimuli. We propose that in such cases, information can be transmitted via spatial spike patterns, employing a sparse or combinatorial coding based on firing rate variance. Around the critical point of a stochastic recurrent excitable network, we uncover a synergistic dual-coding scheme, enabled by single-cell threshold adaptation. This scheme optimizes variance coding for weak signals without compromising rate coding for stronger inputs, thus maximizing input/output mutual information. These optimizations are robust across adaptation rules and coupling strengths through self-suppression of internal noise, particularly around the network's phase transition, and are linked to threshold recovery times observed in hippocampal memory circuits (~$10^2$-$10^3$ms). In contrast, nonadaptive networks perform similarly only at criticality, suggesting that threshold adaptation is essential for reliable encoding of weak signals into diverse spatial patterns. Our results imply a fundamental role for near-critical latent adaptive dynamics enabled by dual coding in biological and artificial neural networks.