Stochastic computing
Arithmetic on random bitstreams — trading precision for tiny, error-tolerant hardware.
Stochastic computing represents a number not as a fixed-point word but as the probability that a bit in a random stream is one. A value of 0.25 is a bitstream that is high a quarter of the time. Under this encoding, multiplication becomes a single AND gate and scaled addition a single multiplexer — arithmetic that would take a full multiplier in binary collapses to a handful of transistors.
The trade is precision for cost and tolerance. A stochastic result has a variance that falls only as the square root of the stream length, so high accuracy is expensive in time; but a bit-flip perturbs the result by one stream position rather than corrupting a high-order bit, which makes the style attractive for noisy, low-power, error-tolerant hardware. It sits naturally alongside spiking and neuromorphic designs, where signals are already event-like.
The neuromorphic studio is built as a universal stochastic-computing framework, and the reproducibility discipline of the whole federation applies to it directly: a probabilistic computation is only auditable if it is deterministic given its seed. The Monte-Carlo lesson makes that concrete with a fixed-seed estimate the reader can reproduce bit-for-bit.
The vocabulary of the topic.
- Bitstream encoding
- A number represented as the probability that a bit in a random stream is one.
- AND-gate multiplication
- Two independent stochastic streams multiplied by a single logic gate.
- Progressive precision
- Accuracy that improves as the stream lengthens; error falls with the square root of length.
- Fault tolerance
- A single bit-flip shifts the value by one stream position, not by a high-order bit.
Studios working on this.
Each runs standalone in its own repository and federates its evidence through the platform.
Run it, don't take it on faith.
The runnable lessons and sealed measurements this site ships for the topic.
Read it at source, in order.
Canonical references for the topic, ordered from the foundations to current work. Every one was verified at source — a DOI resolves through doi.org, a standard through its issuer — so each link goes to the real record.
Start with the foundations
- 1
The founding treatment: numbers as probabilities carried by random bitstreams.
Current work
- 2
The modern survey: how arithmetic on bitstreams trades precision for tiny, error-tolerant hardware.
Related topics.
Where this topic connects to the rest of the federation's work.