| name | stm-sequence-timing-replay |
| description | Spiking Temporal Memory (sTM) model for learning sequence timing and controlling replay speed via oscillatory background inputs. Provides biologically plausible mechanisms for encoding element-specific timing and flexible speed control. |
| version | 1.0.0 |
| author | Melissa Lober, Younes Bouhadjar, Markus Diesmann, Tom Tetzlaff |
| arxiv_id | 2605.22523 |
| created | 2026-05-30T00:00:00.000Z |
| category | neuroscience |
| tags | ["spiking neural network","sequence timing","replay","temporal memory","oscillation","neuroscience"] |
| activation_keywords | ["sequence timing","replay speed","sTM","spiking temporal memory","oscillatory control","temporal encoding"] |
Spiking Temporal Memory (sTM): Sequence Timing and Replay Speed Control
Overview
A biologically inspired network model that extends the spiking Temporal Memory (sTM) framework to learn not only the order of sequence elements but also their precise timing, with flexible control of replay speed via oscillatory background inputs.
Source: arXiv:2605.22523 (Submitted 21 May 2026)
Authors: Melissa Lober, Younes Bouhadjar, Markus Diesmann, Tom Tetzlaff
Category: Quantitative Biology - Neurons and Cognition (q-bio.NC)
Key Concepts
1. Sequence Processing Challenge
- Traditional models learn order but not timing of sequence elements
- Need biologically plausible mechanisms for:
- Encoding element-specific timing
- Flexibly controlling replay speed
2. sTM Model Foundation
- Each sequence element represented by small set of synchronously firing neurons
- Active neuron set encodes element identity in its sequential context
- Original version: learns order but not timing
3. Duration Encoding Mechanism
Element Duration → Sequential Activation of Element-Specific Populations
→ Enables encoding across wide range of timescales
- Duration of each sequence element encoded by:
- Sequential activation of element-specific neuronal populations
- Unique and sparse spatiotemporal patterns for elapsed time
4. Oscillatory Background as Clock Signal
- Oscillatory background inputs serve as clock signal
- Provides robust and flexible mechanism for:
- Controlling sequence replay speed
- Modulating speed during wakefulness and sleep
5. Speed-Replay Correlation
- Replay speed correlates with:
- Characteristics of global oscillatory activity
- Observed in EEG or LFP recordings
- Different during wakefulness vs. sleep
Implementation Approach
Network Architecture
- Element representation: Synchronous firing of small neuron groups
- Timing encoding: Sequential population activation for each element
