[Paper Review] The Synthesis and Analysis of Stochastic Switching Circuits

Hongchao Zhou, Po‐Ling Loh+1 more|arXiv (Cornell University)|Sep 4, 2012

DNA and Biological Computing15 references9 citations

TL;DR

This paper investigates the robustness, expressibility, and approximation of stochastic switching circuits using pswitches with rational probabilities. It proves that simple series-parallel (ssp) circuits are robust to small error perturbations, while general series-parallel circuits are not, and establishes that ssp circuits can synthesize all rational probabilities with denominator $ q^n $ when $ q $ is a multiple of 2 or 3, but not when $ q $ is a prime > 3. A greedy local optimization algorithm is proposed for approximating arbitrary probabilities with exponentially decreasing error as circuit size increases.

ABSTRACT

Stochastic switching circuits are relay circuits that consist of stochastic switches called pswitches. The study of stochastic switching circuits has widespread applications in many fields of computer science, neuroscience, and biochemistry. In this paper, we discuss several properties of stochastic switching circuits, including robustness, expressibility, and probability approximation. First, we study the robustness, namely, the effect caused by introducing an error of size εto each pswitch in a stochastic circuit. We analyze two constructions and prove that simple series-parallel circuits are robust to small error perturbations, while general series-parallel circuits are not. Specifically, the total error introduced by perturbations of size less than εis bounded by a constant multiple of εin a simple series-parallel circuit, independent of the size of the circuit. Next, we study the expressibility of stochastic switching circuits: Given an integer q and a pswitch set S=\{\frac{1}{q},\frac{2}{q},...,\frac{q-1}{q}\}, can we synthesize any rational probability with denominator q^n (for arbitrary n) with a simple series-parallel stochastic switching circuit? We generalize previous results and prove that when q is a multiple of 2 or 3, the answer is yes. We also show that when q is a prime number larger than 3, the answer is no. Probability approximation is studied for a general case of an arbitrary pswitch set S=\{s_1,s_2,...,s_{|S|}\}. In this case, we propose an algorithm based on local optimization to approximate any desired probability. The analysis reveals that the approximation error of a switching circuit decreases exponentially with an increasing circuit size.

Motivation & Objective

• To analyze the robustness of stochastic switching circuits under small error perturbations in pswitch probabilities.
• To determine the set of probabilities that can be exactly synthesized using simple series-parallel (ssp) circuits with a given pswitch set $ S = \{\frac{1}{q}, \frac{2}{q}, \dots, \frac{q-1}{q}\} $.
• To develop an efficient algorithm for approximating arbitrary target probabilities using minimal circuit size and low error.
• To understand the conditions under which ssp circuits can achieve high-precision probability synthesis or approximation.

Proposed method

• Analyzes error propagation in series-parallel circuits by bounding the total error introduced by individual pswitch perturbations of size $ \epsilon $, showing ssp circuits maintain error bounded by a constant multiple of $ \epsilon $, independent of circuit size.
• Uses induction and recursive circuit composition to prove that ssp circuits can realize any rational probability $ \frac{a}{q^n} $ when $ q $ is divisible by 2 or 3, based on the closure probability $ P(C) $ of the circuit.
• Proposes a greedy local optimization algorithm that inserts two pswitches at a time to minimize the approximation error at each step, with insertion rules based on the current probability estimate and the pswitch set $ S $.
• Derives upper bounds on the approximation error using recursive error functions, showing exponential decay with increasing circuit size $ n $, particularly for $ m=2 $ insertions.
• Applies the algorithm to specific cases such as $ S = \{\frac{1}{5}, \frac{2}{5}, \frac{3}{5}, \frac{4}{5}\} $, demonstrating a circuit with 5 pswitches achieving $ \left| \frac{3}{7} - 0.4278 \right| = 7.3 \times 10^{-4} $ error.
• Uses mathematical induction and case analysis to prove that the error reduction factor per step is at most $ \Delta(1 - \Delta) $, where $ \Delta = \frac{1}{q} $, leading to exponential error decay.

Experimental results

Research questions

• RQ1 Under what conditions is a stochastic switching circuit robust to small perturbations in pswitch probabilities?
• RQ2 For a given pswitch set $ S = \{\frac{1}{q}, \frac{2}{q}, \dots, \frac{q-1}{q}\} $, which rational probabilities can be exactly synthesized using simple series-parallel circuits?
• RQ3 Can a general pswitch set be used to approximate any desired probability with low error and small circuit complexity?
• RQ4 How does the approximation error scale with circuit size when using a greedy construction algorithm?
• RQ5 What is the theoretical limit on the approximation accuracy of stochastic switching circuits for arbitrary pswitch sets?

Key findings

• Simple series-parallel (ssp) circuits are robust to small error perturbations: the total error introduced by individual pswitch errors of size $ \epsilon $ is bounded by a constant multiple of $ \epsilon $, independent of circuit size.
• General series-parallel circuits are not robust; error can accumulate and grow with circuit size, even under small individual perturbations.
• When $ q $ is a multiple of 2 or 3, all rational probabilities of the form $ \frac{a}{q^n} $ can be exactly synthesized using ssp circuits with pswitches from $ S = \{\frac{1}{q}, \frac{2}{q}, \dots, \frac{q-1}{q}\} $.
• When $ q $ is a prime number greater than 3, it is impossible to synthesize all such probabilities using ssp circuits, indicating a fundamental limitation in expressibility.
• The greedy algorithm with $ m=2 $ insertions achieves an approximation error upper bound of $ \frac{\Delta}{2} \left( \Delta(1 - \Delta) \right)^{\lceil n/2 \rceil - 1} $, showing exponential decay with circuit size $ n $.
• In a concrete example with $ S = \{\frac{1}{5}, \frac{2}{5}, \frac{3}{5}, \frac{4}{5}\} $, a 5-pswitch circuit approximated $ \frac{3}{7} $ with an error of only $ 7.3 \times 10^{-4} $, confirming the algorithm’s high accuracy.