Disclaimer

NeuroSim is an early-stage software product under active development. It has not been through formal verification or validation, and it may contain errors in its numerical methods, its circuit generators, its measurement code or its presentation of results.

It is provided without any commitment to availability, correctness, stability or continued existence.

The platform is intended for computational neuroscience research, education and evaluation. It is not suitable, and must not be used, as a basis for clinical, diagnostic, regulatory or safety-critical decision-making.

Nothing on this website constitutes medical advice, and nothing here should be used to diagnose, treat, cure or prevent any condition in any organism, human or otherwise. If you need medical advice, consult a qualified clinician.

The simulator does not measure a biological system. It integrates a mathematical model whose parameters were chosen from published literature and whose structure was generated from published architectural descriptions.

Output from this platform is not experimental data, is not evidence about any animal, and must not be reported as an observation of biology.

NeuroSim contains experimental computational models derived from published scientific literature. Alsadaany Industries makes no warranty that those models are scientifically correct, that they faithfully represent the circuits they are named after, or that conclusions drawn from them would survive comparison with experiment.

The neuron model is leaky integrate-and-fire, a first-order abstraction. It does not represent dendritic computation, active conductances, short-term plasticity, neuromodulation, gap junctions, or the electrotonic structure of real neurons.

The published FlyWire connectome of the adult Drosophila brain contains 139,255 neurons and roughly 54.5 million synapses. NeuroSim does not simulate it and does not claim to.

The circuits on this site contain hundreds to a few thousand neurons and are generated from published cell counts, projection rules and connectivity statistics. They are models of described architecture, not copies of measured connectome data. An individual neuron in a NeuroSim circuit does not correspond to an individual biological cell or to any identifier in any connectome dataset.

Integration is numerical and uses a finite timestep. Synaptic weights are derived from a single free parameter of 0.275 mV per synapse. Random circuit structure, membrane noise and Poisson input are pseudo-random draws from a seeded generator. Fixed-point arithmetic is used in the GPU delivery buffer.

Each of these introduces error. A single run is one sample; differences smaller than the run-to-run spread across seeds should not be interpreted as effects.

Connectome datasets, published neuron models, cell-type annotations and scientific software referenced on this site are the work of their respective authors, are governed by their own licences, and are not products of Alsadaany Industries. Their authors bear no responsibility for this implementation or its output. See the third-party notices.

Performance and, in edge cases, numerical results depend on your browser, your operating system, your GPU drivers and your hardware. The WebGPU backend uses different arithmetic ordering and fixed-point accumulation than the CPU reference, and is enabled only after it matches the reference to within a stated tolerance on a probe run.

Frame timing, speed-relative-to-biology figures and wall-clock measurements are properties of your device, not of the model.

Alsadaany Industries claims no scientific discovery, no novel result and no validated finding arising from the platform. Where the site describes a phenomenon such as sparse coding, attractor dynamics or direction selectivity, it is describing published science, with citations, that NeuroSim reproduces in reduced form.