Knowledge

Shannon's source coding theorem

Source 📝

2701: 246:, which quantifies the minimal description length of an object, is more suitable to describe the limits of data compression. Shannon entropy takes into account only frequency regularities while Kolmogorov complexity takes into account all algorithmic regularities, so in general the latter is smaller. On the other hand, if an object is generated by a random process in such a way that it has only frequency regularities, entropy is close to complexity with high probability (Shen et al. 2017). 2208: 42: 3354: 2696:{\displaystyle {\begin{aligned}H(X)&=-\sum _{i=1}^{n}p_{i}\log _{2}p_{i}\\&\leq -\sum _{i=1}^{n}p_{i}\log _{2}q_{i}\\&=-\sum _{i=1}^{n}p_{i}\log _{2}a^{-s_{i}}+\sum _{i=1}^{n}p_{i}\log _{2}C\\&=-\sum _{i=1}^{n}p_{i}\log _{2}a^{-s_{i}}+\log _{2}C\\&\leq -\sum _{i=1}^{n}-s_{i}p_{i}\log _{2}a\\&=\mathbb {E} S\log _{2}a\\\end{aligned}}} 3564: 1204: 223:
data tends to infinity, it is impossible to compress such data such that the code rate (average number of bits per symbol) is less than the Shannon entropy of the source, without it being virtually certain that information will be lost. However it is possible to get the code rate arbitrarily close to
3100: 363:
coded sequence represents the compressed message in a biunivocal way, under the assumption that the decoder knows the source. From a practical point of view, this hypothesis is not always true. Consequently, when the entropy encoding is applied the transmitted message is
1502: 3391: 2024: 1832: 690: 261:
to a sequence of alphabet symbols (usually bits) such that the source symbols can be exactly recovered from the binary bits (lossless source coding) or recovered within some distortion (lossy source coding). This is one approach to
1022: 3349:{\displaystyle {\begin{aligned}\mathbb {E} S&=\sum p_{i}s_{i}\\&<\sum p_{i}\left(-\log _{a}p_{i}+1\right)\\&=\sum -p_{i}{\frac {\log _{2}p_{i}}{\log _{2}a}}+1\\&={\frac {H(X)}{\log _{2}a}}+1\\\end{aligned}}} 1673: 1327: 2029:
The encoding algorithm: the encoder checks if the input sequence lies within the typical set; if yes, it outputs the index of the input sequence within the typical set; if not, the encoder outputs an arbitrary
2961: 3664: 3888: 3105: 2864: 2213: 2785: 991: 3085: 1897: 1738: 1571: 281: 3017: 1368: 2179: 433: 734: 220: 201: 1353: 3559:{\displaystyle A_{n}^{\varepsilon }=\left\{x_{1}^{n}\ :\ \left|-{\frac {1}{n}}\log p\left(X_{1},\cdots ,X_{n}\right)-{\overline {H_{n}}}(X)\right|<\varepsilon \right\}.} 715: 1922: 3770: 1745: 361: 580: 1199:{\displaystyle A_{n}^{\varepsilon }=\left\{(x_{1},\cdots ,x_{n})\ :\ \left|-{\frac {1}{n}}\log p(x_{1},\cdots ,x_{n})-H_{n}(X)\right|<\varepsilon \right\}.} 164: 1601: 3790: 3601:. Now we just encode the sequences in the typical set, and usual methods in source coding show that the cardinality of this set is smaller than 3883: 1240: 96: 274:
In information theory, the source coding theorem (Shannon 1948) informally states that (MacKay 2003, pg. 81, Cover 2006, Chapter 5):
3846: 3823: 3754: 2875: 1356: 1210: 107: 157: 3604: 81: 3799: 2807: 3723: 757: 232: 133: 51: 2720: 867: 138: 3868: 3028: 150: 112: 435:. Usually, the information that characterizes the source is inserted at the beginning of the transmitted message. 3878: 1837: 1678: 177:
This article is about the theory of source coding in data compression. For the term in computer programming, see
1514: 1497:{\displaystyle 2^{-n(H(X)+\varepsilon )}\leq p\left(x_{1},\cdots ,x_{n}\right)\leq 2^{-n(H(X)-\varepsilon )}} 231:
places an upper and a lower bound on the minimal possible expected length of codewords as a function of the
3873: 2972: 3090:
and so by Kraft's inequality there exists a prefix-free code having those word lengths. Thus the minimal
242:
Note that, for data that exhibits more dependencies (whose source is not an i.i.d. random variable), the
2126: 2056:), the encoder does not make any error. So, the probability of error of the encoder is bounded above by 367: 243: 3743:
Shen, A. and Uspensky, V.A. and Vereshchagin, N. (2017). "Chapter 7.3. : Complexity and entropy".
2711: 2707: 771: 505: 76: 56: 1332: 3787: 2049:
digit number. As long as the input sequence lies within the typical set (with probability at least
2019:{\displaystyle \left|A_{n}^{\varepsilon }\right|\leq 2^{n(H(X)+\varepsilon )},n(H(X)+\varepsilon )} 61: 698: 3764: 1362:
The definition of typical sets implies that those sequences that lie in the typical set satisfy:
782: 258: 185: 71: 33: 3842: 3819: 3750: 1827:{\displaystyle \left|A_{n}^{\varepsilon }\right|\geq (1-\varepsilon )2^{n(H(X)-\varepsilon )}} 3814: 263: 197: 128: 86: 685:{\displaystyle {\frac {H(X)}{\log _{2}a}}\leq \mathbb {E} <{\frac {H(X)}{\log _{2}a}}+1} 334: 3794: 236: 205: 3718: 3713: 718: 212: 3365:
Fixed rate lossless source coding for discrete time non-stationary independent sources
3862: 66: 17: 41: 3783: 1217:, the probability that a sequence generated by the source lies in the typical set, 91: 3836: 3744: 997: 738: 479: 178: 2082:(in the sense of exponent) would cover a set of probability bounded away from 1668:{\displaystyle \left|A_{n}^{\varepsilon }\right|\leq 2^{n(H(X)+\varepsilon )}} 774:
in the continuous-valued case. The Source coding theorem states that for any
570:
is optimal in the sense that it has the minimal expected word length for
799: 2066:: the converse is proved by showing that any set of size smaller than 1322:{\displaystyle P((X_{1},X_{2},\cdots ,X_{n})\in A_{n}^{\varepsilon })} 1329:
can be made arbitrarily close to 1, and specifically, greater than
1233:, as defined approaches one. In particular, for sufficiently large 841:
are recoverable from the binary bits with probability of at least
221:
independent and identically-distributed random variable (i.i.d.)
2956:{\displaystyle -\log _{a}p_{i}\leq s_{i}<-\log _{a}p_{i}+1} 1899:
and the lower bound on the total probability of the whole set
306: 257:
is a mapping from (a sequence of) symbols from an information
547:
denote the random variable given by the length of codeword
327:
bits it is virtually certain that information will be lost.
3659:{\displaystyle 2^{n({\overline {H_{n}}}(X)+\varepsilon )}} 1675:, which follows from the left hand side (lower bound) for 224:
the Shannon entropy, with negligible probability of loss.
3803:, vol. 27, pp. 379–423, 623-656, July, October, 1948 316:; but conversely, if they are compressed into fewer than 3835:
Cover, Thomas M. (2006). "Chapter 5: Data Compression".
3685:
bits suffice for encoding with probability greater than
219:
shows that, in the limit, as the length of a stream of
3815:
Information Theory, Inference, and Learning Algorithms
3889:
Mathematical theorems in theoretical computer science
3607: 3394: 3103: 3031: 2975: 2878: 2810: 2723: 2211: 2129: 1925: 1840: 1748: 1681: 1604: 1517: 1371: 1335: 1243: 1025: 870: 701: 583: 370: 337: 2026:
bits are enough to point to any string in this set.
2859:{\displaystyle s_{i}=\lceil -\log _{a}p_{i}\rceil } 202:
independent identically-distributed random variable
27:
Establishes the limits to possible data compression
3658: 3558: 3348: 3079: 3011: 2955: 2858: 2779: 2695: 2173: 2018: 1891: 1826: 1732: 1667: 1565: 1496: 1347: 1321: 1198: 985: 709: 684: 427: 355: 2780:{\displaystyle C=\sum _{i=1}^{n}a^{-s_{i}}\leq 1} 986:{\displaystyle p(x_{1},\ldots ,x_{n})=\Pr \left.} 196:) establishes the statistical limits to possible 3746:Kolmogorov Complexity and Algorithmic Randomness 3080:{\displaystyle \sum a^{-s_{i}}\leq \sum p_{i}=1} 912: 3360:Extension to non-stationary independent sources 3749:. American Mathematical Society. p. 226. 3818:Cambridge: Cambridge University Press, 2003. 2090:Proof: Source coding theorem for symbol codes 309:with negligible risk of information loss, as 158: 8: 3769:: CS1 maint: multiple names: authors list ( 2853: 2824: 3841:. John Wiley & Sons. pp. 103–142. 1892:{\displaystyle p(x_{1},x_{2},\cdots x_{n})} 1733:{\displaystyle p(x_{1},x_{2},\cdots x_{n})} 1566:{\displaystyle (X_{1},X_{2},\cdots X_{n})} 239:) and of the size of the target alphabet. 165: 151: 29: 3700:can be made arbitrarily small, by making 3625: 3619: 3612: 3606: 3517: 3511: 3497: 3478: 3450: 3427: 3422: 3404: 3399: 3393: 3321: 3300: 3269: 3257: 3244: 3237: 3231: 3194: 3181: 3163: 3140: 3130: 3109: 3108: 3104: 3102: 3065: 3047: 3039: 3030: 3003: 2988: 2980: 2974: 2941: 2928: 2912: 2899: 2886: 2877: 2847: 2834: 2815: 2809: 2763: 2755: 2745: 2734: 2722: 2677: 2666: 2665: 2643: 2633: 2623: 2610: 2599: 2570: 2555: 2547: 2534: 2524: 2514: 2503: 2474: 2464: 2454: 2443: 2428: 2420: 2407: 2397: 2387: 2376: 2353: 2340: 2330: 2320: 2309: 2286: 2273: 2263: 2253: 2242: 2212: 2210: 2163: 2155: 2147: 2134: 2128: 1956: 1939: 1934: 1924: 1880: 1864: 1851: 1839: 1794: 1762: 1757: 1747: 1721: 1705: 1692: 1680: 1635: 1618: 1613: 1603: 1554: 1538: 1525: 1516: 1461: 1443: 1424: 1376: 1370: 1334: 1310: 1305: 1289: 1270: 1257: 1242: 1162: 1146: 1127: 1101: 1075: 1056: 1035: 1030: 1024: 969: 956: 937: 924: 900: 881: 869: 835:binary bits such that the source symbols 703: 702: 700: 661: 640: 624: 623: 605: 584: 582: 369: 336: 2114:denote the word length of each possible 235:of the input word (which is viewed as a 3735: 32: 3788:A Mathematical Theory of Communication 3762: 489:is a random variable taking values in 439:Source coding theorem for symbol codes 229:source coding theorem for symbol codes 2801:For the second inequality we may set 1834:, which follows from upper bound for 802:of the source, there is large enough 482:from those alphabets (respectively). 204:, and the operational meaning of the 7: 3012:{\displaystyle a^{-s_{i}}\leq p_{i}} 454:denote two finite alphabets and let 2706:where the second line follows from 284:random variables each with entropy 97:Limiting density of discrete points 2174:{\displaystyle q_{i}=a^{-s_{i}}/C} 1213:(AEP) shows that for large enough 428:{\displaystyle NH(X)+(inf.source)} 25: 1211:asymptotic equipartition property 810:i.i.d. repetition of the source, 295:can be compressed into more than 108:Asymptotic equipartition property 2710:and the fifth line follows from 770:in the discrete-valued case and 40: 190:Shannon's source coding theorem 124:Shannon's source coding theorem 3838:Elements of Information Theory 3651: 3642: 3636: 3616: 3534: 3528: 3312: 3306: 2225: 2219: 2013: 2004: 1998: 1992: 1981: 1972: 1966: 1960: 1886: 1844: 1819: 1810: 1804: 1798: 1787: 1775: 1727: 1685: 1660: 1651: 1645: 1639: 1560: 1518: 1511:The probability of a sequence 1489: 1480: 1474: 1468: 1404: 1395: 1389: 1383: 1348:{\displaystyle 1-\varepsilon } 1316: 1295: 1250: 1247: 1174: 1168: 1152: 1120: 1081: 1049: 906: 874: 652: 646: 634: 628: 596: 590: 422: 389: 383: 377: 350: 344: 82:Conditional mutual information 1: 3800:Bell System Technical Journal 3884:Presentation layer protocols 3724:Noisy-channel coding theorem 3631: 3523: 725:Proof: source coding theorem 710:{\displaystyle \mathbb {E} } 200:for data whose source is an 134:Noisy-channel coding theorem 3905: 806:and an encoder that takes 176: 1016:, is defined as follows: 194:noiseless coding theorem 3666:. Thus, on an average, 852:Proof of Achievability. 574:, then (Shannon 1948): 480:set of all finite words 139:Shannon–Hartley theorem 3660: 3560: 3350: 3081: 3013: 2957: 2860: 2781: 2750: 2697: 2615: 2519: 2459: 2392: 2325: 2258: 2175: 2020: 1893: 1828: 1734: 1669: 1567: 1498: 1349: 1323: 1200: 987: 711: 686: 429: 357: 329: 113:Rate–distortion theory 3661: 3561: 3351: 3082: 3014: 2958: 2861: 2782: 2730: 2698: 2595: 2499: 2439: 2372: 2305: 2238: 2176: 2021: 1894: 1829: 1735: 1670: 1568: 1499: 1350: 1324: 1201: 988: 712: 687: 430: 358: 356:{\displaystyle NH(X)} 276: 270:Source coding theorem 244:Kolmogorov complexity 217:source coding theorem 18:Source coding theorem 3812:David J. C. MacKay. 3605: 3392: 3101: 3029: 2973: 2876: 2808: 2721: 2209: 2127: 1923: 1838: 1746: 1679: 1602: 1515: 1369: 1333: 1241: 1023: 868: 772:differential entropy 699: 581: 368: 335: 77:Directed information 57:Differential entropy 3432: 3409: 3369:Define typical set 1944: 1767: 1623: 1315: 1040: 62:Conditional entropy 3869:Information theory 3793:2009-02-16 at the 3656: 3556: 3418: 3395: 3346: 3344: 3077: 3009: 2953: 2856: 2777: 2712:Kraft's inequality 2693: 2691: 2185:is chosen so that 2171: 2016: 1930: 1889: 1824: 1753: 1730: 1665: 1609: 1563: 1494: 1345: 1319: 1301: 1196: 1026: 983: 707: 682: 506:uniquely decodable 425: 353: 186:information theory 72:Mutual information 34:Information theory 3634: 3526: 3458: 3441: 3435: 3334: 3282: 2708:Gibbs' inequality 2064:Proof of converse 1573:being drawn from 1109: 1092: 1086: 816:, and maps it to 674: 618: 175: 174: 16:(Redirected from 3896: 3879:Data compression 3853: 3852: 3832: 3826: 3810: 3804: 3781: 3775: 3774: 3768: 3760: 3740: 3703: 3699: 3695: 3691: 3684: 3675: 3665: 3663: 3662: 3657: 3655: 3654: 3635: 3630: 3629: 3620: 3600: 3595: 3594: 3579: 3575: 3569:Then, for given 3565: 3563: 3562: 3557: 3552: 3548: 3541: 3537: 3527: 3522: 3521: 3512: 3507: 3503: 3502: 3501: 3483: 3482: 3459: 3451: 3439: 3433: 3431: 3426: 3408: 3403: 3384: 3383: 3382: 3355: 3353: 3352: 3347: 3345: 3335: 3333: 3326: 3325: 3315: 3301: 3293: 3283: 3281: 3274: 3273: 3263: 3262: 3261: 3249: 3248: 3238: 3236: 3235: 3214: 3210: 3206: 3199: 3198: 3186: 3185: 3168: 3167: 3149: 3145: 3144: 3135: 3134: 3112: 3093: 3086: 3084: 3083: 3078: 3070: 3069: 3054: 3053: 3052: 3051: 3018: 3016: 3015: 3010: 3008: 3007: 2995: 2994: 2993: 2992: 2962: 2960: 2959: 2954: 2946: 2945: 2933: 2932: 2917: 2916: 2904: 2903: 2891: 2890: 2865: 2863: 2862: 2857: 2852: 2851: 2839: 2838: 2820: 2819: 2797: 2786: 2784: 2783: 2778: 2770: 2769: 2768: 2767: 2749: 2744: 2702: 2700: 2699: 2694: 2692: 2682: 2681: 2669: 2658: 2648: 2647: 2638: 2637: 2628: 2627: 2614: 2609: 2585: 2575: 2574: 2562: 2561: 2560: 2559: 2539: 2538: 2529: 2528: 2518: 2513: 2489: 2479: 2478: 2469: 2468: 2458: 2453: 2435: 2434: 2433: 2432: 2412: 2411: 2402: 2401: 2391: 2386: 2362: 2358: 2357: 2345: 2344: 2335: 2334: 2324: 2319: 2295: 2291: 2290: 2278: 2277: 2268: 2267: 2257: 2252: 2201: 2184: 2180: 2178: 2177: 2172: 2167: 2162: 2161: 2160: 2159: 2139: 2138: 2122: 2113: 2104: 2085: 2081: 2080: 2079: 2059: 2055: 2048: 2025: 2023: 2022: 2017: 1985: 1984: 1948: 1943: 1938: 1914: 1913: 1912: 1898: 1896: 1895: 1890: 1885: 1884: 1869: 1868: 1856: 1855: 1833: 1831: 1830: 1825: 1823: 1822: 1771: 1766: 1761: 1739: 1737: 1736: 1731: 1726: 1725: 1710: 1709: 1697: 1696: 1674: 1672: 1671: 1666: 1664: 1663: 1627: 1622: 1617: 1595: 1589:is greater than 1588: 1587: 1586: 1572: 1570: 1569: 1564: 1559: 1558: 1543: 1542: 1530: 1529: 1503: 1501: 1500: 1495: 1493: 1492: 1453: 1449: 1448: 1447: 1429: 1428: 1408: 1407: 1354: 1352: 1351: 1346: 1328: 1326: 1325: 1320: 1314: 1309: 1294: 1293: 1275: 1274: 1262: 1261: 1236: 1232: 1231: 1230: 1216: 1205: 1203: 1202: 1197: 1192: 1188: 1181: 1177: 1167: 1166: 1151: 1150: 1132: 1131: 1110: 1102: 1090: 1084: 1080: 1079: 1061: 1060: 1039: 1034: 1015: 1014: 1013: 992: 990: 989: 984: 979: 975: 974: 973: 961: 960: 942: 941: 929: 928: 905: 904: 886: 885: 860: 847: 840: 834: 815: 809: 805: 798:larger than the 797: 780: 769: 755: 732: 716: 714: 713: 708: 706: 691: 689: 688: 683: 675: 673: 666: 665: 655: 641: 627: 619: 617: 610: 609: 599: 585: 573: 569: 558: 546: 542: 531: 530: 529: 519: 518: 517: 503: 495: 488: 477: 476: 475: 465: 464: 463: 453: 434: 432: 431: 426: 362: 360: 359: 354: 326: 315: 305: 294: 280: 264:data compression 198:data compression 167: 160: 153: 129:Channel capacity 87:Relative entropy 44: 30: 21: 3904: 3903: 3899: 3898: 3897: 3895: 3894: 3893: 3859: 3858: 3857: 3856: 3849: 3834: 3833: 3829: 3811: 3807: 3795:Wayback Machine 3782: 3778: 3761: 3757: 3742: 3741: 3737: 3732: 3710: 3701: 3697: 3693: 3686: 3673: 3668: 3667: 3621: 3608: 3603: 3602: 3593: 3588: 3587: 3586: 3581: 3577: 3570: 3513: 3493: 3474: 3473: 3469: 3446: 3442: 3417: 3413: 3390: 3389: 3381: 3376: 3375: 3374: 3370: 3367: 3362: 3343: 3342: 3317: 3316: 3302: 3291: 3290: 3265: 3264: 3253: 3240: 3239: 3227: 3212: 3211: 3190: 3177: 3173: 3169: 3159: 3147: 3146: 3136: 3126: 3116: 3099: 3098: 3091: 3061: 3043: 3035: 3027: 3026: 2999: 2984: 2976: 2971: 2970: 2937: 2924: 2908: 2895: 2882: 2874: 2873: 2843: 2830: 2811: 2806: 2805: 2791: 2759: 2751: 2719: 2718: 2690: 2689: 2673: 2656: 2655: 2639: 2629: 2619: 2583: 2582: 2566: 2551: 2543: 2530: 2520: 2487: 2486: 2470: 2460: 2424: 2416: 2403: 2393: 2360: 2359: 2349: 2336: 2326: 2293: 2292: 2282: 2269: 2259: 2228: 2207: 2206: 2198: 2192: 2186: 2182: 2151: 2143: 2130: 2125: 2124: 2120: 2115: 2111: 2106: 2095: 2092: 2083: 2078: 2073: 2072: 2071: 2067: 2057: 2050: 2031: 1952: 1926: 1921: 1920: 1911: 1906: 1905: 1904: 1900: 1876: 1860: 1847: 1836: 1835: 1790: 1749: 1744: 1743: 1717: 1701: 1688: 1677: 1676: 1631: 1605: 1600: 1599: 1590: 1585: 1580: 1579: 1578: 1574: 1550: 1534: 1521: 1513: 1512: 1507: 1457: 1439: 1420: 1419: 1415: 1372: 1367: 1366: 1331: 1330: 1285: 1266: 1253: 1239: 1238: 1234: 1229: 1224: 1223: 1222: 1218: 1214: 1158: 1142: 1123: 1097: 1093: 1071: 1052: 1048: 1044: 1021: 1020: 1012: 1007: 1006: 1005: 1001: 965: 952: 933: 920: 919: 915: 896: 877: 866: 865: 855: 842: 836: 817: 811: 807: 803: 785: 781:, i.e. for any 775: 760: 756:is i.i.d. with 753: 747: 741: 730: 727: 697: 696: 657: 656: 642: 601: 600: 586: 579: 578: 571: 563: 548: 544: 537: 533: 528: 525: 524: 523: 521: 516: 513: 512: 511: 509: 497: 494: 490: 486: 474: 471: 470: 469: 467: 462: 459: 458: 457: 455: 452: 448: 444: 441: 366: 365: 333: 332: 317: 310: 296: 285: 278: 272: 252: 237:random variable 206:Shannon entropy 182: 171: 28: 23: 22: 15: 12: 11: 5: 3902: 3900: 3892: 3891: 3886: 3881: 3876: 3871: 3861: 3860: 3855: 3854: 3847: 3827: 3805: 3776: 3755: 3734: 3733: 3731: 3728: 3727: 3726: 3721: 3719:Error exponent 3716: 3714:Channel coding 3709: 3706: 3671: 3653: 3650: 3647: 3644: 3641: 3638: 3633: 3628: 3624: 3618: 3615: 3611: 3589: 3580:large enough, 3567: 3566: 3555: 3551: 3547: 3544: 3540: 3536: 3533: 3530: 3525: 3520: 3516: 3510: 3506: 3500: 3496: 3492: 3489: 3486: 3481: 3477: 3472: 3468: 3465: 3462: 3457: 3454: 3449: 3445: 3438: 3430: 3425: 3421: 3416: 3412: 3407: 3402: 3398: 3377: 3366: 3363: 3361: 3358: 3357: 3356: 3341: 3338: 3332: 3329: 3324: 3320: 3314: 3311: 3308: 3305: 3299: 3296: 3294: 3292: 3289: 3286: 3280: 3277: 3272: 3268: 3260: 3256: 3252: 3247: 3243: 3234: 3230: 3226: 3223: 3220: 3217: 3215: 3213: 3209: 3205: 3202: 3197: 3193: 3189: 3184: 3180: 3176: 3172: 3166: 3162: 3158: 3155: 3152: 3150: 3148: 3143: 3139: 3133: 3129: 3125: 3122: 3119: 3117: 3115: 3111: 3107: 3106: 3088: 3087: 3076: 3073: 3068: 3064: 3060: 3057: 3050: 3046: 3042: 3038: 3034: 3020: 3019: 3006: 3002: 2998: 2991: 2987: 2983: 2979: 2964: 2963: 2952: 2949: 2944: 2940: 2936: 2931: 2927: 2923: 2920: 2915: 2911: 2907: 2902: 2898: 2894: 2889: 2885: 2881: 2867: 2866: 2855: 2850: 2846: 2842: 2837: 2833: 2829: 2826: 2823: 2818: 2814: 2788: 2787: 2776: 2773: 2766: 2762: 2758: 2754: 2748: 2743: 2740: 2737: 2733: 2729: 2726: 2704: 2703: 2688: 2685: 2680: 2676: 2672: 2668: 2664: 2661: 2659: 2657: 2654: 2651: 2646: 2642: 2636: 2632: 2626: 2622: 2618: 2613: 2608: 2605: 2602: 2598: 2594: 2591: 2588: 2586: 2584: 2581: 2578: 2573: 2569: 2565: 2558: 2554: 2550: 2546: 2542: 2537: 2533: 2527: 2523: 2517: 2512: 2509: 2506: 2502: 2498: 2495: 2492: 2490: 2488: 2485: 2482: 2477: 2473: 2467: 2463: 2457: 2452: 2449: 2446: 2442: 2438: 2431: 2427: 2423: 2419: 2415: 2410: 2406: 2400: 2396: 2390: 2385: 2382: 2379: 2375: 2371: 2368: 2365: 2363: 2361: 2356: 2352: 2348: 2343: 2339: 2333: 2329: 2323: 2318: 2315: 2312: 2308: 2304: 2301: 2298: 2296: 2294: 2289: 2285: 2281: 2276: 2272: 2266: 2262: 2256: 2251: 2248: 2245: 2241: 2237: 2234: 2231: 2229: 2227: 2224: 2221: 2218: 2215: 2214: 2196: 2190: 2170: 2166: 2158: 2154: 2150: 2146: 2142: 2137: 2133: 2118: 2109: 2091: 2088: 2074: 2015: 2012: 2009: 2006: 2003: 2000: 1997: 1994: 1991: 1988: 1983: 1980: 1977: 1974: 1971: 1968: 1965: 1962: 1959: 1955: 1951: 1947: 1942: 1937: 1933: 1929: 1917: 1916: 1907: 1888: 1883: 1879: 1875: 1872: 1867: 1863: 1859: 1854: 1850: 1846: 1843: 1821: 1818: 1815: 1812: 1809: 1806: 1803: 1800: 1797: 1793: 1789: 1786: 1783: 1780: 1777: 1774: 1770: 1765: 1760: 1756: 1752: 1741: 1729: 1724: 1720: 1716: 1713: 1708: 1704: 1700: 1695: 1691: 1687: 1684: 1662: 1659: 1656: 1653: 1650: 1647: 1644: 1641: 1638: 1634: 1630: 1626: 1621: 1616: 1612: 1608: 1597: 1581: 1562: 1557: 1553: 1549: 1546: 1541: 1537: 1533: 1528: 1524: 1520: 1505: 1504: 1491: 1488: 1485: 1482: 1479: 1476: 1473: 1470: 1467: 1464: 1460: 1456: 1452: 1446: 1442: 1438: 1435: 1432: 1427: 1423: 1418: 1414: 1411: 1406: 1403: 1400: 1397: 1394: 1391: 1388: 1385: 1382: 1379: 1375: 1359:for a proof). 1344: 1341: 1338: 1318: 1313: 1308: 1304: 1300: 1297: 1292: 1288: 1284: 1281: 1278: 1273: 1269: 1265: 1260: 1256: 1252: 1249: 1246: 1225: 1207: 1206: 1195: 1191: 1187: 1184: 1180: 1176: 1173: 1170: 1165: 1161: 1157: 1154: 1149: 1145: 1141: 1138: 1135: 1130: 1126: 1122: 1119: 1116: 1113: 1108: 1105: 1100: 1096: 1089: 1083: 1078: 1074: 1070: 1067: 1064: 1059: 1055: 1051: 1047: 1043: 1038: 1033: 1029: 1008: 994: 993: 982: 978: 972: 968: 964: 959: 955: 951: 948: 945: 940: 936: 932: 927: 923: 918: 914: 911: 908: 903: 899: 895: 892: 889: 884: 880: 876: 873: 751: 745: 726: 723: 719:expected value 705: 693: 692: 681: 678: 672: 669: 664: 660: 654: 651: 648: 645: 639: 636: 633: 630: 626: 622: 616: 613: 608: 604: 598: 595: 592: 589: 535: 526: 514: 492: 472: 460: 450: 446: 440: 437: 424: 421: 418: 415: 412: 409: 406: 403: 400: 397: 394: 391: 388: 385: 382: 379: 376: 373: 352: 349: 346: 343: 340: 271: 268: 251: 248: 213:Claude Shannon 173: 172: 170: 169: 162: 155: 147: 144: 143: 142: 141: 136: 131: 126: 118: 117: 116: 115: 110: 102: 101: 100: 99: 94: 89: 84: 79: 74: 69: 64: 59: 54: 46: 45: 37: 36: 26: 24: 14: 13: 10: 9: 6: 4: 3: 2: 3901: 3890: 3887: 3885: 3882: 3880: 3877: 3875: 3874:Coding theory 3872: 3870: 3867: 3866: 3864: 3850: 3848:0-471-24195-4 3844: 3840: 3839: 3831: 3828: 3825: 3824:0-521-64298-1 3821: 3817: 3816: 3809: 3806: 3802: 3801: 3796: 3792: 3789: 3785: 3780: 3777: 3772: 3766: 3758: 3756:9781470431822 3752: 3748: 3747: 3739: 3736: 3729: 3725: 3722: 3720: 3717: 3715: 3712: 3711: 3707: 3705: 3690: 3683: 3679: 3674: 3648: 3645: 3639: 3626: 3622: 3613: 3609: 3599: 3592: 3585: 3573: 3553: 3549: 3545: 3542: 3538: 3531: 3518: 3514: 3508: 3504: 3498: 3494: 3490: 3487: 3484: 3479: 3475: 3470: 3466: 3463: 3460: 3455: 3452: 3447: 3443: 3436: 3428: 3423: 3419: 3414: 3410: 3405: 3400: 3396: 3388: 3387: 3386: 3380: 3373: 3364: 3359: 3339: 3336: 3330: 3327: 3322: 3318: 3309: 3303: 3297: 3295: 3287: 3284: 3278: 3275: 3270: 3266: 3258: 3254: 3250: 3245: 3241: 3232: 3228: 3224: 3221: 3218: 3216: 3207: 3203: 3200: 3195: 3191: 3187: 3182: 3178: 3174: 3170: 3164: 3160: 3156: 3153: 3151: 3141: 3137: 3131: 3127: 3123: 3120: 3118: 3113: 3097: 3096: 3095: 3074: 3071: 3066: 3062: 3058: 3055: 3048: 3044: 3040: 3036: 3032: 3025: 3024: 3023: 3004: 3000: 2996: 2989: 2985: 2981: 2977: 2969: 2968: 2967: 2950: 2947: 2942: 2938: 2934: 2929: 2925: 2921: 2918: 2913: 2909: 2905: 2900: 2896: 2892: 2887: 2883: 2879: 2872: 2871: 2870: 2848: 2844: 2840: 2835: 2831: 2827: 2821: 2816: 2812: 2804: 2803: 2802: 2799: 2795: 2774: 2771: 2764: 2760: 2756: 2752: 2746: 2741: 2738: 2735: 2731: 2727: 2724: 2717: 2716: 2715: 2713: 2709: 2686: 2683: 2678: 2674: 2670: 2662: 2660: 2652: 2649: 2644: 2640: 2634: 2630: 2624: 2620: 2616: 2611: 2606: 2603: 2600: 2596: 2592: 2589: 2587: 2579: 2576: 2571: 2567: 2563: 2556: 2552: 2548: 2544: 2540: 2535: 2531: 2525: 2521: 2515: 2510: 2507: 2504: 2500: 2496: 2493: 2491: 2483: 2480: 2475: 2471: 2465: 2461: 2455: 2450: 2447: 2444: 2440: 2436: 2429: 2425: 2421: 2417: 2413: 2408: 2404: 2398: 2394: 2388: 2383: 2380: 2377: 2373: 2369: 2366: 2364: 2354: 2350: 2346: 2341: 2337: 2331: 2327: 2321: 2316: 2313: 2310: 2306: 2302: 2299: 2297: 2287: 2283: 2279: 2274: 2270: 2264: 2260: 2254: 2249: 2246: 2243: 2239: 2235: 2232: 2230: 2222: 2216: 2205: 2204: 2203: 2199: 2189: 2168: 2164: 2156: 2152: 2148: 2144: 2140: 2135: 2131: 2121: 2112: 2103: 2099: 2089: 2087: 2077: 2070: 2065: 2061: 2054: 2046: 2042: 2038: 2034: 2027: 2010: 2007: 2001: 1995: 1989: 1986: 1978: 1975: 1969: 1963: 1957: 1953: 1949: 1945: 1940: 1935: 1931: 1927: 1910: 1903: 1881: 1877: 1873: 1870: 1865: 1861: 1857: 1852: 1848: 1841: 1816: 1813: 1807: 1801: 1795: 1791: 1784: 1781: 1778: 1772: 1768: 1763: 1758: 1754: 1750: 1742: 1722: 1718: 1714: 1711: 1706: 1702: 1698: 1693: 1689: 1682: 1657: 1654: 1648: 1642: 1636: 1632: 1628: 1624: 1619: 1614: 1610: 1606: 1598: 1594: 1584: 1577: 1555: 1551: 1547: 1544: 1539: 1535: 1531: 1526: 1522: 1510: 1509: 1508: 1486: 1483: 1477: 1471: 1465: 1462: 1458: 1454: 1450: 1444: 1440: 1436: 1433: 1430: 1425: 1421: 1416: 1412: 1409: 1401: 1398: 1392: 1386: 1380: 1377: 1373: 1365: 1364: 1363: 1360: 1358: 1342: 1339: 1336: 1311: 1306: 1302: 1298: 1290: 1286: 1282: 1279: 1276: 1271: 1267: 1263: 1258: 1254: 1244: 1228: 1221: 1212: 1193: 1189: 1185: 1182: 1178: 1171: 1163: 1159: 1155: 1147: 1143: 1139: 1136: 1133: 1128: 1124: 1117: 1114: 1111: 1106: 1103: 1098: 1094: 1087: 1076: 1072: 1068: 1065: 1062: 1057: 1053: 1045: 1041: 1036: 1031: 1027: 1019: 1018: 1017: 1011: 1004: 999: 980: 976: 970: 966: 962: 957: 953: 949: 946: 943: 938: 934: 930: 925: 921: 916: 909: 901: 897: 893: 890: 887: 882: 878: 871: 864: 863: 862: 858: 853: 849: 846: 839: 832: 828: 824: 820: 814: 801: 796: 792: 788: 784: 778: 773: 767: 763: 759: 754: 744: 740: 736: 724: 722: 720: 679: 676: 670: 667: 662: 658: 649: 643: 637: 631: 620: 614: 611: 606: 602: 593: 587: 577: 576: 575: 567: 560: 556: 552: 541: 507: 501: 485:Suppose that 483: 481: 438: 436: 419: 416: 413: 410: 407: 404: 401: 398: 395: 392: 386: 380: 374: 371: 347: 341: 338: 328: 324: 320: 313: 308: 303: 299: 292: 288: 283: 275: 269: 267: 265: 260: 256: 255:Source coding 249: 247: 245: 240: 238: 234: 230: 225: 222: 218: 214: 209: 207: 203: 199: 195: 191: 187: 180: 168: 163: 161: 156: 154: 149: 148: 146: 145: 140: 137: 135: 132: 130: 127: 125: 122: 121: 120: 119: 114: 111: 109: 106: 105: 104: 103: 98: 95: 93: 90: 88: 85: 83: 80: 78: 75: 73: 70: 68: 67:Joint entropy 65: 63: 60: 58: 55: 53: 50: 49: 48: 47: 43: 39: 38: 35: 31: 19: 3837: 3830: 3813: 3808: 3798: 3784:C.E. Shannon 3779: 3745: 3738: 3688: 3681: 3677: 3669: 3597: 3590: 3583: 3571: 3568: 3378: 3371: 3368: 3089: 3021: 2965: 2868: 2800: 2793: 2789: 2705: 2194: 2187: 2116: 2107: 2101: 2097: 2093: 2075: 2068: 2063: 2062: 2052: 2044: 2040: 2036: 2032: 2028: 1918: 1908: 1901: 1592: 1582: 1575: 1506: 1361: 1226: 1219: 1208: 1009: 1002: 995: 856: 851: 850: 844: 837: 830: 826: 822: 818: 812: 794: 790: 786: 776: 765: 761: 749: 742: 737:source, its 728: 717:denotes the 694: 565: 561: 554: 550: 539: 499: 484: 442: 330: 322: 318: 311: 301: 297: 290: 286: 277: 273: 254: 253: 241: 228: 226: 216: 211:Named after 210: 193: 189: 183: 123: 92:Entropy rate 3596:) > 1 − 998:typical set 739:time series 478:denote the 179:Source code 3863:Categories 3730:References 3094:satisfies 861:, and let 721:operator. 508:code from 250:Statements 3765:cite book 3649:ε 3632:¯ 3546:ε 3524:¯ 3509:− 3488:⋯ 3464:⁡ 3448:− 3406:ε 3328:⁡ 3276:⁡ 3251:⁡ 3225:− 3222:∑ 3188:⁡ 3175:− 3157:∑ 3124:∑ 3059:∑ 3056:≤ 3041:− 3033:∑ 2997:≤ 2982:− 2935:⁡ 2922:− 2906:≤ 2893:⁡ 2880:− 2854:⌉ 2841:⁡ 2828:− 2825:⌈ 2772:≤ 2757:− 2732:∑ 2684:⁡ 2650:⁡ 2617:− 2597:∑ 2593:− 2590:≤ 2577:⁡ 2549:− 2541:⁡ 2501:∑ 2497:− 2481:⁡ 2441:∑ 2422:− 2414:⁡ 2374:∑ 2370:− 2347:⁡ 2307:∑ 2303:− 2300:≤ 2280:⁡ 2240:∑ 2236:− 2149:− 2123:. Define 2011:ε 1979:ε 1950:≤ 1941:ε 1874:⋯ 1817:ε 1814:− 1785:ε 1782:− 1773:≥ 1764:ε 1715:⋯ 1658:ε 1629:≤ 1620:ε 1548:⋯ 1487:ε 1484:− 1463:− 1455:≤ 1434:⋯ 1410:≤ 1402:ε 1378:− 1343:ε 1340:− 1312:ε 1299:∈ 1280:⋯ 1186:ε 1156:− 1137:⋯ 1115:⁡ 1099:− 1066:⋯ 1037:ε 947:⋯ 891:… 854:Fix some 668:⁡ 621:≤ 612:⁡ 319:N H 298:N H 3791:Archived 3708:See also 3704:larger. 3692:, where 2869:so that 2202:. Then 2193:+ ... + 2181:, where 553: ( 496:and let 2966:and so 800:entropy 758:entropy 748:, ..., 568:  564:  549:  502:  498:  233:entropy 52:Entropy 3845:  3822:  3753:  3576:, for 3574:> 0 3440:  3434:  1919:Since 1355:(See 1091:  1085:  859:> 0 779:> 0 735:i.i.d. 733:is an 729:Given 695:Where 543:. Let 532:where 282:i.i.d. 259:source 215:, the 504:be a 3843:ISBN 3820:ISBN 3771:link 3751:ISBN 3696:and 3687:1 − 3680:) + 3543:< 3385:as: 3154:< 3022:and 2919:< 2792:log 2105:let 2096:1 ≤ 2094:For 2051:1 − 2043:) + 1591:1 − 1209:The 1183:< 996:The 843:1 − 829:) + 793:) + 783:rate 638:< 538:| = 466:and 443:Let 331:The 307:bits 227:The 192:(or 3797:", 3786:, " 3582:Pr( 3461:log 3319:log 3267:log 3242:log 3179:log 2926:log 2884:log 2832:log 2796:≤ 0 2790:so 2675:log 2641:log 2568:log 2532:log 2472:log 2405:log 2338:log 2271:log 2200:= 1 1357:AEP 1112:log 659:log 603:log 562:If 520:to 449:, Σ 314:→ ∞ 184:In 3865:: 3767:}} 3763:{{ 2798:. 2714:: 2100:≤ 2086:. 2060:. 1237:, 1000:, 913:Pr 848:. 559:. 534:|Σ 266:. 208:. 188:, 3851:. 3773:) 3759:. 3702:n 3698:δ 3694:ε 3689:δ 3682:ε 3678:X 3676:( 3672:n 3670:H 3652:) 3646:+ 3643:) 3640:X 3637:( 3627:n 3623:H 3617:( 3614:n 3610:2 3598:δ 3591:n 3584:A 3578:n 3572:δ 3554:. 3550:} 3539:| 3535:) 3532:X 3529:( 3519:n 3515:H 3505:) 3499:n 3495:X 3491:, 3485:, 3480:1 3476:X 3471:( 3467:p 3456:n 3453:1 3444:| 3437:: 3429:n 3424:1 3420:x 3415:{ 3411:= 3401:n 3397:A 3379:n 3372:A 3340:1 3337:+ 3331:a 3323:2 3313:) 3310:X 3307:( 3304:H 3298:= 3288:1 3285:+ 3279:a 3271:2 3259:i 3255:p 3246:2 3233:i 3229:p 3219:= 3208:) 3204:1 3201:+ 3196:i 3192:p 3183:a 3171:( 3165:i 3161:p 3142:i 3138:s 3132:i 3128:p 3121:= 3114:S 3110:E 3092:S 3075:1 3072:= 3067:i 3063:p 3049:i 3045:s 3037:a 3005:i 3001:p 2990:i 2986:s 2978:a 2951:1 2948:+ 2943:i 2939:p 2930:a 2914:i 2910:s 2901:i 2897:p 2888:a 2849:i 2845:p 2836:a 2822:= 2817:i 2813:s 2794:C 2775:1 2765:i 2761:s 2753:a 2747:n 2742:1 2739:= 2736:i 2728:= 2725:C 2687:a 2679:2 2671:S 2667:E 2663:= 2653:a 2645:2 2635:i 2631:p 2625:i 2621:s 2612:n 2607:1 2604:= 2601:i 2580:C 2572:2 2564:+ 2557:i 2553:s 2545:a 2536:2 2526:i 2522:p 2516:n 2511:1 2508:= 2505:i 2494:= 2484:C 2476:2 2466:i 2462:p 2456:n 2451:1 2448:= 2445:i 2437:+ 2430:i 2426:s 2418:a 2409:2 2399:i 2395:p 2389:n 2384:1 2381:= 2378:i 2367:= 2355:i 2351:q 2342:2 2332:i 2328:p 2322:n 2317:1 2314:= 2311:i 2288:i 2284:p 2275:2 2265:i 2261:p 2255:n 2250:1 2247:= 2244:i 2233:= 2226:) 2223:X 2220:( 2217:H 2197:n 2195:q 2191:1 2188:q 2183:C 2169:C 2165:/ 2157:i 2153:s 2145:a 2141:= 2136:i 2132:q 2119:i 2117:x 2110:i 2108:s 2102:n 2098:i 2084:1 2076:n 2069:A 2058:ε 2053:ε 2047:) 2045:ε 2041:X 2039:( 2037:H 2035:( 2033:n 2014:) 2008:+ 2005:) 2002:X 1999:( 1996:H 1993:( 1990:n 1987:, 1982:) 1976:+ 1973:) 1970:X 1967:( 1964:H 1961:( 1958:n 1954:2 1946:| 1936:n 1932:A 1928:| 1915:. 1909:n 1902:A 1887:) 1882:n 1878:x 1871:, 1866:2 1862:x 1858:, 1853:1 1849:x 1845:( 1842:p 1820:) 1811:) 1808:X 1805:( 1802:H 1799:( 1796:n 1792:2 1788:) 1779:1 1776:( 1769:| 1759:n 1755:A 1751:| 1740:. 1728:) 1723:n 1719:x 1712:, 1707:2 1703:x 1699:, 1694:1 1690:x 1686:( 1683:p 1661:) 1655:+ 1652:) 1649:X 1646:( 1643:H 1640:( 1637:n 1633:2 1625:| 1615:n 1611:A 1607:| 1596:. 1593:ε 1583:n 1576:A 1561:) 1556:n 1552:X 1545:, 1540:2 1536:X 1532:, 1527:1 1523:X 1519:( 1490:) 1481:) 1478:X 1475:( 1472:H 1469:( 1466:n 1459:2 1451:) 1445:n 1441:x 1437:, 1431:, 1426:1 1422:x 1417:( 1413:p 1405:) 1399:+ 1396:) 1393:X 1390:( 1387:H 1384:( 1381:n 1374:2 1337:1 1317:) 1307:n 1303:A 1296:) 1291:n 1287:X 1283:, 1277:, 1272:2 1268:X 1264:, 1259:1 1255:X 1251:( 1248:( 1245:P 1235:n 1227:n 1220:A 1215:n 1194:. 1190:} 1179:| 1175:) 1172:X 1169:( 1164:n 1160:H 1153:) 1148:n 1144:x 1140:, 1134:, 1129:1 1125:x 1121:( 1118:p 1107:n 1104:1 1095:| 1088:: 1082:) 1077:n 1073:x 1069:, 1063:, 1058:1 1054:x 1050:( 1046:{ 1042:= 1032:n 1028:A 1010:n 1003:A 981:. 977:] 971:n 967:x 963:= 958:n 954:X 950:, 944:, 939:1 935:x 931:= 926:1 922:X 917:[ 910:= 907:) 902:n 898:x 894:, 888:, 883:1 879:x 875:( 872:p 857:ε 845:ε 838:X 833:) 831:ε 827:X 825:( 823:H 821:( 819:n 813:X 808:n 804:n 795:ε 791:X 789:( 787:H 777:ε 768:) 766:X 764:( 762:H 752:n 750:X 746:1 743:X 731:X 704:E 680:1 677:+ 671:a 663:2 653:) 650:X 647:( 644:H 635:] 632:S 629:[ 625:E 615:a 607:2 597:) 594:X 591:( 588:H 572:X 566:f 557:) 555:X 551:f 545:S 540:a 536:2 527:2 522:Σ 515:1 510:Σ 500:f 493:1 491:Σ 487:X 473:2 468:Σ 461:1 456:Σ 451:2 447:1 445:Σ 423:) 420:e 417:c 414:r 411:u 408:o 405:s 402:. 399:f 396:n 393:i 390:( 387:+ 384:) 381:X 378:( 375:H 372:N 351:) 348:X 345:( 342:H 339:N 325:) 323:X 321:( 312:N 304:) 302:X 300:( 293:) 291:X 289:( 287:H 279:N 181:. 166:e 159:t 152:v 20:)

Index

Source coding theorem
Information theory

Entropy
Differential entropy
Conditional entropy
Joint entropy
Mutual information
Directed information
Conditional mutual information
Relative entropy
Entropy rate
Limiting density of discrete points
Asymptotic equipartition property
Rate–distortion theory
Shannon's source coding theorem
Channel capacity
Noisy-channel coding theorem
Shannon–Hartley theorem
v
t
e
Source code
information theory
data compression
independent identically-distributed random variable
Shannon entropy
Claude Shannon
independent and identically-distributed random variable (i.i.d.)
entropy

Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.