Skip to main content
Log in

On the Reliability Function for a BSC with Noiseless Feedback at Zero Rate

  • INFORMATION THEORY
  • Published:
Problems of Information Transmission Aims and scope Submit manuscript

Abstract

We consider the transmission of nonexponentially many messages through a binary symmetric channel with noiseless feedback. We obtain an upper bound for the best decoding error exponent. Combined with the corresponding known lower bound, this allows us to find the reliability function for this channel at zero rate.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1.

Similar content being viewed by others

References

  1. Berlekamp, E.R., Block Coding with Noiseless Feedback, PhD Thesis, MIT, Cambridge, USA, 1964. Available at http://hdl.handle.net/1721.1/14783.

  2. Zigangirov, K.Sh., Upper Bounds for the Error Probability for Channels with Feedback, Probl. Peredachi Inf., 1970, vol. 6, no. 2, pp. 87–92 [Probl. Inf. Transm. (Engl. Transl.), 1970, vol. 6, no. 2, pp. 159–163]. http://mi.mathnet.ru/eng/ppi1740

    MathSciNet  Google Scholar 

  3. Burnashev, M.V., On the Reliability Function of a Binary Symmetrical Channel with Feedback, Probl. Peredachi Inf., 1988, vol. 24, no. 1, pp. 3–10 [Probl. Inf. Transm. (Engl. Transl.), 1988, vol. 24, no. 1, pp. 1–7]. http://mi.mathnet.ru/eng/ppi681

    MathSciNet  MATH  Google Scholar 

  4. Zigangirov, K.Sh., Optimum Zero Rate Transmission through Binary Symmetric Channel with Feedback, Probl. Control Inform. Theory, 1978, vol. 7, no. 3, pp. 183–198.

    MathSciNet  MATH  Google Scholar 

  5. Shannon, C.E., Probability of Error for Optimal Codes in a Gaussian Channel, Bell Syst. Tech. J., 1959, vol. 38, no. 3, pp. 611–656. https://doi.org/10.1002/j.1538-7305.1959.tb03905.x

    Article  MathSciNet  Google Scholar 

  6. Gallager, R.G., Information Theory and Reliable Communication, New York: Wiley, 1968. Translated under the title Teoriya informatsii i nadezhnaya svyaz’, Moscow: Sov. Radio, 1974.

    MATH  Google Scholar 

  7. Pinsker, M.S., The Probability of Error in Block Transmission in a Memoryless Gaussian Channel with Feedback, Probl. Peredachi Inf., 1968, vol. 4, no. 4, pp. 3–19 [Probl. Inf. Transm. (Engl. Transl.), 1968, vol. 4, no. 4, pp. 1–14]. http://mi.mathnet.ru/eng/ppi1868

    MathSciNet  Google Scholar 

  8. Korn, G.A. and Korn, T.M., Mathematical Handbook for Scientists and Engineers, New York: McGraw-Hill, 1968. Translated under the title Spravochnik po matematike dlya nauchnykh rabotnikov i inzhenerov, Moscow: Nauka, 1974.

    MATH  Google Scholar 

Download references

Acknowledgments

The author is grateful to L.A. Bassalygo and G.A. Kabatiansky for useful discussions and constructive critical remarks, which improved the paper.

Funding

Supported in part by the Russian Foundation for Basic Research, project no. 19-01-00364.

Author information

Authors and Affiliations

Authors

Additional information

Translated from Problemy Peredachi Informatsii, 2022, Vol. 58, No. 3, pp. 3–17. https://doi.org/10.31857/S0555292322030019

Appendix

Appendix

Proof of equation (14).

Consider an \(n\)-simplex code \((\boldsymbol{x}_1,\boldsymbol{x}_2,\boldsymbol{x}_3)\) with \(\boldsymbol{x}_1\) having first \(n/3\) ones and then \(2n/3\) zeros; \(\boldsymbol{x}_2\) having first \(n/3\) zeros, then \(n/3\) ones, and then \(n/3\) zeros; and \(\boldsymbol{x}_3\) having first \(2n/3\) zeros and then \(n/3\) ones. Then \(w(\boldsymbol{x}_1)=w(\boldsymbol{x}_2)=w(\boldsymbol{x}_3)=n/3\) and \(d_{12}=d_{13}=d_{23}=2n/3\). Assume that an output \(\boldsymbol{y}\) has \(u_1n/3\) ones in the first \(n/3\) positions, \(u_2n/3\) ones in the next \(n/3\) positions, and \(u_3n/3\) ones in the last \(n/3\) positions. Then

$$\begin{gathered}d(\boldsymbol{x}_1,\boldsymbol{y})/n=(1-u_1+u_2+u_3)/3,\\ d(\boldsymbol{x}_2,\boldsymbol{y})/n=(1+u_1-u_2+u_3)/3,\\ d(\boldsymbol{x}_3,\boldsymbol{y})/n=(1+u_1+u_2-u_3)/3.\end{gathered}$$

Since \(d(\boldsymbol{x}_1,\boldsymbol{y})=d(\boldsymbol{x}_2,\boldsymbol{y})=d(\boldsymbol{x}_3,\boldsymbol{y})\), we obtain \(u_1=u_2=u_3\) and

$$p(\boldsymbol{y}^n\,|\, \boldsymbol{x}_1)=p^{d(\boldsymbol{x}_1,\boldsymbol{y})} q^{n-d(\boldsymbol{x}_1,\boldsymbol{y})}=q^n z^{d(\boldsymbol{x}_1,\boldsymbol{y})}= q^nz^{(1+u)n/3},\quad z=p/q<1.$$

Therefore,

$$ \begin{aligned} \operatorname{\mathbf P}\nolimits\bigl\{p(\boldsymbol{y}^n\,|\, \boldsymbol{x}_1) \approx p(\boldsymbol{y}^n\,|\, \boldsymbol{x}_2) \approx p(\boldsymbol{y}^n\,|\, \boldsymbol{x}_3)\bigr\} &\sim \max_{0\le u\le 1}\operatorname{\mathbf P}\nolimits\Bigl\{p(\boldsymbol{y}^n\,|\, \boldsymbol{x}_1) \approx q^nz^{(1+u)n/3}\Bigr\}\\ &\sim \max_{0\le u\le 1} \biggl\{\binom{n}{un}p^{(1+u)n/3}q^{(2-u)n/3}\biggr\}\\ &\sim q^n\max_{0\le u\le 1}\biggl\{\binom{n}{un}z^{(1+u)n/3}\biggr\}, \end{aligned}$$

and

$$\frac{1}{n}\max_{0\le u\le 1}\ln\operatorname{\mathbf P}\nolimits\{p(\boldsymbol{y}^n\,|\, \boldsymbol{x}_1)\}=\ln q+\max_{0\le u\le 1}g(u),$$
(59)

where

$$g(u)=h(u)+(1+u)\ln(z^{1/3}),\qquad g'(u)=\ln\frac{1-u}{u}+\ln(z^{1/3}),\qquad g''(u)<0.$$

For a maximizing \(u_0\), we obtain

$$ u_0 =\frac{1}{1+z^{-1/3}}= \frac{p^{1/3}}{p^{1/3}+ q^{1/3}},$$

and after a simple algebra,

$$\ln q+g(u_0)=\ln\bigl(p^{1/3}q^{2/3}+p^{2/3}q^{1/3}\bigr).$$
(60)

Now (59) and (60) imply equations (14) and (15). △

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Burnashev, M. On the Reliability Function for a BSC with Noiseless Feedback at Zero Rate. Probl Inf Transm 58, 203–216 (2022). https://doi.org/10.1134/S0032946022030012

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1134/S0032946022030012

Keywords

Navigation