Twitch Neural Network Model for Neuromorphic Ballistic Control in Pong

A research paper by Jordi Timmermans, uploaded to figshare on 2026-04 13, describing a Twitch Neural Network model for neuromuscular-inspired ballistic control. The 1.4 MB PDF details a system where spiking neural network outputs generate muscle-twitch signals to control an actuator in an adapted Pong game, achieving hit rates up to 99%.

Use Cases

• Benchmarking neuromorphic control algorithms based on the described Twitch Neural Network architecture.
• Studying the application of Direct Feedback Alignment learning rules for spiking networks in a control task.
• Investigating the robustness of spiking systems to membrane potential noise during inference, as analyzed in the paper.
• Developing actuator control systems inspired by biological sensorimotor pathways, as described in the work.

Strengths

• Model performance is quantified with specific hit rates (96%, 98%, 99%) under different configurations.
• Noise and threshold robustness are analyzed with concrete parameters (e.g., uniform noise up to 0.1-1% of neuron threshold).
• The dataset is openly licensed under CC-BY-4.0.

Limitations

• The primary file is a 1.4 MB PDF; underlying simulation data, code, or column-level details are not provided.
• Row count and data structure are unknown, limiting suitability assessment for direct ML training.
• The description focuses on model methodology; the availability of raw training or inference logs is unclear.

Provenance

Source

figshare, authored by Jordi Timmermans.

Collection Method

Research output describing a computational model and its training results.

Time Range

Publication date is 2026-04-13.

Freshness

Last updated 2026-04-13 05:45:14.

Geography

Geographic scope is not specified.