History & Evolution

Probabilistic and stochastic computing is not new — it resurfaces whenever energy, reliability, or massive parallelism outrank cycle-by-cycle exactness.

timeline
    title Stochastic & Probabilistic Computing Timeline
    1960s : Gaines — stochastic computing systems
    2000s : Brown & Card — digital SC revival for VLSI
    2010s : Alaghi / Hayes / Qian — synthesis, correlation, DSP/NN apps
    2015+ : Edge AI, LDPC, deterministic SC, Sobol SNGs
    2020s : p-bits, Ising machines, SC compute-in-memory

---

1960s–1970s: Probability as a number format

Early SC treated a wire’s duty cycle (or Bernoulli rate) as the numeric value. Multiplication, scaling, and some nonlinearities collapsed to a few gates. The approach faded as binary VLSI scaled aggressively under Dennard scaling — exact wide datapaths became “free enough.”

---

2000s–2010s: VLSI revival

As power density and dark silicon rose, researchers revisited SC for:

• Image processing kernels that tolerate density noise
• Neural network inference with tiny MAC units
• CAD / synthesis of stochastic circuits (spectral methods, correlation-aware design)

Surveys and tutorials from this era remain the best entry points; look them up by DOI via the Paper Map and classic IEEE search terms (stochastic computing, stochastic logic).

---

Mid-2010s onward: Determinism, coding, and learning

Three threads dominate modern literature (reflected in this portal’s taxonomies):

1. Better streams — low-discrepancy and deterministic bitstreams shrink \(N\) for a target MSE.
2. Application engines — stochastic LDPC / MIMO detectors and CNN / BNN accelerators.
3. Beyond classical SC — p-bits and invertible networks for combinatorial optimization and sampling.

---

Relation to inexact / approximate computing

Idea	Approximate binary	Stochastic / probabilistic
Error model	Truncation / logic approximation	Sampling variance + correlation bias
Typical unit	Approx adder / multiplier cell	Gate + SNG + counter
Knob	Bit significance, compressor truth table	Stream length \(N\), RNG quality

Both families serve error-resilient apps; they are complementary. See also the Inexact Computing portal.

---

📚 Pointers (DOI only)

Foundational and survey-style entry points (publisher pages; no PDFs hosted here):

• B. R. Gaines, early stochastic computing systems literature (search IEEE / historical reprints).
• B. D. Brown and H. C. Card, Stochastic Neural Computation, IEEE Trans. Computers — DOI 10.1109/12.954505.
• A. Alaghi and J. P. Hayes, Survey of Stochastic Computing, ACM TECS — DOI 10.1145/2465787.2465794.
• W. Qian et al., An Architecture for Fault-Tolerant Computation with Stochastic Logic, IEEE Trans. Computers — DOI 10.1109/TC.2010.202.