Knowledge

Stress majorization

Source 📝

1287: 1533: 995: 2137: 2036: 537: 1660: 2345: 1897: 1282:{\displaystyle \sigma (X)=\sum _{i<j\leq n}w_{ij}(d_{ij}(X)-\delta _{ij})^{2}=\sum _{i<j}w_{ij}\delta _{ij}^{2}+\sum _{i<j}w_{ij}d_{ij}^{2}(X)-2\sum _{i<j}w_{ij}\delta _{ij}d_{ij}(X)} 2273: 1395: 2363:. That is, one can find a reasonably aesthetically appealing layout for a network or graph by minimizing a stress function over the positions of the nodes in the graph. In this case, the 2503: 1717: 2214: 1813: 2391: 721: 576: 291: 1942: 875: 823: 196: 2549: 1387: 647: 1756: 2523: 987: 927: 907: 388: 162: 2465: 2175: 771: 608: 2042: 1565: 360: 337: 2434: 2412: 1353: 1330: 1310: 947: 843: 794: 741: 691: 671: 412: 313: 258: 238: 217: 134: 113: 92: 70: 49: 1950: 773:
can be used to specify a degree of confidence in the similarity between points (e.g. 0 can be specified if there is no information for a particular pair).
424: 2601:
de Leeuw, J. (1977), "Applications of convex analysis to multidimensional scaling", in Barra, J. R.; Brodeau, F.; Romie, G.; et al. (eds.),
881:
approach. However, a significantly better (in terms of guarantees on, and rate of, convergence) method for minimizing stress was introduced by
961:
function. This iterative majorization process is also referred to as the SMACOF algorithm ("Scaling by MAjorizing a COmplicated Function").
1573: 2675: 2280: 1820: 1528:{\displaystyle \sum _{i<j}w_{ij}\delta _{ij}d_{ij}(X)=\,\operatorname {tr} \,X'B(X)X\geq \,\operatorname {tr} \,X'B(Z)Z} 2220: 2735: 825:
gives a plot in which points that are close together correspond to points that are also close together in the original
2525:
is chosen as a trade-off between preserving long- or short-range ideal distances. Good results have been shown for
2740: 2470: 1665: 2730: 2180: 1761: 23: 2651: 27: 2366: 696: 545: 264: 1912: 851: 799: 172: 2725: 2646: 2528: 613: 1725: 2508: 2132:{\displaystyle \leq C+\,\operatorname {tr} \,X'VX-2\,\operatorname {tr} \,X'B(Z)Z=\tau (X,Z)} 972: 912: 892: 373: 138: 2571:(1964), "Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis", 2440: 2150: 746: 2692:
Cohen, J. (1997), "Drawing graphs to convey proximity: an incremental arrangement method",
2359:
Stress majorization and algorithms similar to SMACOF also have application in the field of
581: 1541: 1361: 8: 2031:{\displaystyle \sigma (X)=C+\,\operatorname {tr} \,X'VX-2\,\operatorname {tr} \,X'B(X)X} 342: 319: 2419: 2397: 1338: 1315: 1295: 932: 828: 779: 726: 676: 656: 650: 397: 298: 243: 223: 202: 119: 98: 77: 55: 34: 2680:, Lecture Notes in Computer Science, vol. 3383, Springer-Verlag, pp. 239–250 2701: 2656: 2580: 878: 2674:
Gansner, E.; Koren, Y.; North, S. (2004), "Graph Drawing by Stress Majorization",
2637:
Michailidis, G.; de Leeuw, J. (2001), "Data visualization through graph drawing",
2618: 954: 532:{\displaystyle \sigma (X)=\sum _{i<j\leq n}w_{ij}(d_{ij}(X)-\delta _{ij})^{2}} 363: 2568: 2351:
This algorithm has been shown to decrease stress monotonically (see de Leeuw).
1903: 1333: 2719: 2360: 391: 1356: 882: 2705: 2660: 1389:) and therefore relatively easily solved. The third term is bounded by: 889:
method at each step minimizes a simple convex function which both bounds
2584: 877:
could be minimized. For example, Kruskal recommended an iterative
723:
is the ideal distance between the points (their separation) in the
367: 2393:
are usually set to the graph-theoretic distances between nodes
1655:{\displaystyle b_{ij}=-{\frac {w_{ij}\delta _{ij}}{d_{ij}(Z)}}} 2340:{\displaystyle \sigma (X^{k-1})-\sigma (X^{k})<\epsilon } 165:-dimensional space is sought that minimizes the so-called 578:
is a weight for the measurement between a pair of points
2623:
Modern Multidimensional Scaling: theory and applications
2531: 2511: 2473: 2443: 2422: 2400: 2369: 2283: 2223: 2183: 2153: 2045: 1953: 1915: 1892:{\displaystyle b_{ii}=-\sum _{j=1,j\neq i}^{n}b_{ij}} 1823: 1764: 1728: 1668: 1576: 1544: 1398: 1364: 1341: 1318: 1298: 998: 975: 935: 915: 895: 854: 831: 802: 782: 749: 729: 699: 679: 659: 616: 584: 548: 427: 400: 394:
that measures the squared differences between ideal (
376: 345: 322: 301: 267: 246: 226: 205: 175: 141: 122: 101: 80: 58: 37: 2677:
Proceedings of 12th Int. Symp. Graph Drawing (GD'04)
2636: 2268:{\displaystyle X^{k}\leftarrow \min _{X}\tau (X,Z)} 2543: 2517: 2497: 2459: 2428: 2406: 2385: 2339: 2267: 2208: 2169: 2131: 2030: 1936: 1891: 1807: 1750: 1711: 1654: 1559: 1527: 1381: 1347: 1324: 1304: 1281: 981: 941: 921: 901: 869: 837: 817: 788: 765: 735: 715: 685: 665: 641: 602: 570: 531: 406: 382: 354: 331: 307: 285: 252: 232: 211: 190: 156: 128: 107: 86: 64: 43: 2673: 2717: 2238: 414:-dimensional) distances and actual distances in 2694:ACM Transactions on Computer-Human Interaction 2143:The iterative minimization procedure is then: 366:so that the result may be visualised (i.e. an 16:Geometric placement based on ideal distances 2616: 2596: 2594: 1909:Thus, we have a simple quadratic function 2650: 2084: 2080: 2059: 2055: 2004: 2000: 1979: 1975: 1906:inequality, see Borg (pp. 152–153). 1501: 1497: 1470: 1466: 73:-dimensional data items, a configuration 2600: 2591: 2567: 2498:{\displaystyle \delta _{ij}^{-\alpha }} 2354: 1712:{\displaystyle d_{ij}(Z)\neq 0,i\neq j} 1292:Note that the first term is a constant 964: 418:-dimensional space. It is defined as: 2718: 909:from above and touches the surface of 2691: 1312:and the second term is quadratic in 743:-dimensional data space. Note that 2610: 2209:{\displaystyle Z\leftarrow X^{k-1}} 1902:Proof of this inequality is by the 1808:{\displaystyle d_{ij}(Z)=0,i\neq j} 13: 14: 2752: 2603:Recent developments in statistics 1355:the second term is equivalent to 2685: 2667: 2630: 2561: 2328: 2315: 2306: 2287: 2262: 2250: 2234: 2187: 2126: 2114: 2102: 2096: 2022: 2016: 1963: 1957: 1931: 1919: 1784: 1778: 1688: 1682: 1646: 1640: 1554: 1548: 1519: 1513: 1488: 1482: 1460: 1454: 1276: 1270: 1206: 1200: 1091: 1071: 1065: 1049: 1008: 1002: 864: 858: 812: 806: 636: 630: 597: 585: 520: 500: 494: 478: 437: 431: 280: 268: 185: 179: 151: 142: 1: 2554: 2386:{\displaystyle \delta _{ij}} 989:can be expanded as follows: 957:such a function is called a 716:{\displaystyle \delta _{ij}} 571:{\displaystyle w_{ij}\geq 0} 7: 2625:, New York: Springer-Verlag 286:{\displaystyle (n\times r)} 10: 2757: 1937:{\displaystyle \tau (X,Z)} 870:{\displaystyle \sigma (X)} 818:{\displaystyle \sigma (X)} 191:{\displaystyle \sigma (X)} 30:(MDS) where, for a set of 2736:Mathematical optimization 2544:{\displaystyle \alpha =2} 848:There are many ways that 845:-dimensional data space. 642:{\displaystyle d_{ij}(X)} 1751:{\displaystyle b_{ij}=0} 28:multidimensional scaling 2518:{\displaystyle \alpha } 1944:that majorizes stress: 982:{\displaystyle \sigma } 922:{\displaystyle \sigma } 902:{\displaystyle \sigma } 383:{\displaystyle \sigma } 157:{\displaystyle (\ll m)} 2545: 2519: 2499: 2461: 2460:{\displaystyle w_{ij}} 2430: 2408: 2387: 2341: 2269: 2210: 2171: 2170:{\displaystyle k^{th}} 2133: 2032: 1938: 1893: 1875: 1809: 1752: 1713: 1656: 1561: 1529: 1383: 1349: 1326: 1306: 1283: 983: 943: 923: 903: 887:iterative majorization 871: 839: 819: 790: 767: 766:{\displaystyle w_{ij}} 737: 717: 687: 667: 643: 604: 572: 533: 408: 384: 356: 333: 309: 287: 254: 234: 213: 192: 158: 130: 109: 88: 66: 45: 2741:Mathematical analysis 2706:10.1145/264645.264657 2661:10.1007/s001800100077 2546: 2520: 2500: 2462: 2431: 2409: 2388: 2342: 2270: 2211: 2172: 2134: 2033: 1939: 1894: 1843: 1810: 1753: 1714: 1657: 1562: 1530: 1384: 1350: 1327: 1307: 1284: 984: 944: 924: 904: 872: 840: 820: 791: 768: 738: 718: 688: 668: 644: 605: 603:{\displaystyle (i,j)} 573: 534: 409: 385: 357: 334: 310: 288: 255: 235: 214: 193: 159: 131: 110: 89: 67: 46: 24:optimization strategy 2529: 2509: 2471: 2441: 2420: 2398: 2367: 2355:Use in graph drawing 2281: 2221: 2181: 2151: 2043: 1951: 1913: 1821: 1762: 1726: 1666: 1574: 1560:{\displaystyle B(Z)} 1542: 1396: 1382:{\displaystyle X'VX} 1362: 1339: 1316: 1296: 996: 973: 969:The stress function 965:The SMACOF algorithm 933: 913: 893: 852: 829: 800: 780: 747: 727: 697: 677: 657: 614: 582: 546: 425: 398: 374: 343: 320: 299: 265: 244: 224: 203: 173: 139: 120: 99: 78: 56: 35: 2731:Dimension reduction 2494: 1199: 1149: 20:Stress majorization 2605:, pp. 133–145 2585:10.1007/BF02289565 2541: 2515: 2495: 2474: 2457: 2426: 2404: 2383: 2337: 2265: 2246: 2206: 2167: 2129: 2028: 1934: 1889: 1805: 1748: 1709: 1652: 1557: 1525: 1414: 1379: 1345: 1322: 1302: 1279: 1230: 1182: 1168: 1132: 1118: 1035: 979: 939: 919: 899: 867: 835: 815: 786: 763: 733: 713: 683: 663: 651:euclidean distance 639: 600: 568: 529: 464: 404: 380: 355:{\displaystyle 3-} 352: 332:{\displaystyle 2-} 329: 305: 283: 250: 230: 209: 188: 154: 126: 105: 84: 62: 41: 2639:Computation Stat. 2429:{\displaystyle j} 2407:{\displaystyle i} 2347:otherwise repeat. 2237: 1650: 1399: 1348:{\displaystyle V} 1325:{\displaystyle X} 1305:{\displaystyle C} 1215: 1153: 1103: 1014: 942:{\displaystyle Z} 838:{\displaystyle m} 789:{\displaystyle X} 736:{\displaystyle m} 686:{\displaystyle j} 666:{\displaystyle i} 443: 407:{\displaystyle m} 370:). The function 308:{\displaystyle X} 253:{\displaystyle 3} 233:{\displaystyle 2} 212:{\displaystyle r} 129:{\displaystyle r} 108:{\displaystyle n} 87:{\displaystyle X} 65:{\displaystyle m} 44:{\displaystyle n} 2748: 2710: 2708: 2689: 2683: 2681: 2671: 2665: 2663: 2654: 2634: 2628: 2626: 2614: 2608: 2606: 2598: 2589: 2587: 2565: 2550: 2548: 2547: 2542: 2524: 2522: 2521: 2516: 2504: 2502: 2501: 2496: 2493: 2485: 2467:are taken to be 2466: 2464: 2463: 2458: 2456: 2455: 2437:and the weights 2435: 2433: 2432: 2427: 2413: 2411: 2410: 2405: 2392: 2390: 2389: 2384: 2382: 2381: 2346: 2344: 2343: 2338: 2327: 2326: 2305: 2304: 2274: 2272: 2271: 2266: 2245: 2233: 2232: 2215: 2213: 2212: 2207: 2205: 2204: 2176: 2174: 2173: 2168: 2166: 2165: 2138: 2136: 2135: 2130: 2092: 2067: 2037: 2035: 2034: 2029: 2012: 1987: 1943: 1941: 1940: 1935: 1898: 1896: 1895: 1890: 1888: 1887: 1874: 1869: 1836: 1835: 1814: 1812: 1811: 1806: 1777: 1776: 1757: 1755: 1754: 1749: 1741: 1740: 1718: 1716: 1715: 1710: 1681: 1680: 1661: 1659: 1658: 1653: 1651: 1649: 1639: 1638: 1625: 1624: 1623: 1611: 1610: 1597: 1589: 1588: 1566: 1564: 1563: 1558: 1534: 1532: 1531: 1526: 1509: 1478: 1453: 1452: 1440: 1439: 1427: 1426: 1413: 1388: 1386: 1385: 1380: 1372: 1354: 1352: 1351: 1346: 1331: 1329: 1328: 1323: 1311: 1309: 1308: 1303: 1288: 1286: 1285: 1280: 1269: 1268: 1256: 1255: 1243: 1242: 1229: 1198: 1193: 1181: 1180: 1167: 1148: 1143: 1131: 1130: 1117: 1099: 1098: 1089: 1088: 1064: 1063: 1048: 1047: 1034: 988: 986: 985: 980: 951:supporting point 948: 946: 945: 940: 928: 926: 925: 920: 908: 906: 905: 900: 879:steepest descent 876: 874: 873: 868: 844: 842: 841: 836: 824: 822: 821: 816: 796:which minimizes 795: 793: 792: 787: 776:A configuration 772: 770: 769: 764: 762: 761: 742: 740: 739: 734: 722: 720: 719: 714: 712: 711: 692: 690: 689: 684: 672: 670: 669: 664: 648: 646: 645: 640: 629: 628: 609: 607: 606: 601: 577: 575: 574: 569: 561: 560: 538: 536: 535: 530: 528: 527: 518: 517: 493: 492: 477: 476: 463: 413: 411: 410: 405: 389: 387: 386: 381: 361: 359: 358: 353: 338: 336: 335: 330: 316:lists points in 314: 312: 311: 306: 292: 290: 289: 284: 259: 257: 256: 251: 239: 237: 236: 231: 218: 216: 215: 210: 197: 195: 194: 189: 163: 161: 160: 155: 135: 133: 132: 127: 114: 112: 111: 106: 93: 91: 90: 85: 71: 69: 68: 63: 50: 48: 47: 42: 2756: 2755: 2751: 2750: 2749: 2747: 2746: 2745: 2716: 2715: 2714: 2713: 2690: 2686: 2672: 2668: 2635: 2631: 2615: 2611: 2599: 2592: 2566: 2562: 2557: 2530: 2527: 2526: 2510: 2507: 2506: 2486: 2478: 2472: 2469: 2468: 2448: 2444: 2442: 2439: 2438: 2421: 2418: 2417: 2399: 2396: 2395: 2374: 2370: 2368: 2365: 2364: 2357: 2322: 2318: 2294: 2290: 2282: 2279: 2278: 2241: 2228: 2224: 2222: 2219: 2218: 2194: 2190: 2182: 2179: 2178: 2158: 2154: 2152: 2149: 2148: 2142: 2085: 2060: 2044: 2041: 2040: 2005: 1980: 1952: 1949: 1948: 1914: 1911: 1910: 1880: 1876: 1870: 1847: 1828: 1824: 1822: 1819: 1818: 1769: 1765: 1763: 1760: 1759: 1733: 1729: 1727: 1724: 1723: 1673: 1669: 1667: 1664: 1663: 1631: 1627: 1626: 1616: 1612: 1603: 1599: 1598: 1596: 1581: 1577: 1575: 1572: 1571: 1543: 1540: 1539: 1502: 1471: 1445: 1441: 1432: 1428: 1419: 1415: 1403: 1397: 1394: 1393: 1365: 1363: 1360: 1359: 1340: 1337: 1336: 1317: 1314: 1313: 1297: 1294: 1293: 1261: 1257: 1248: 1244: 1235: 1231: 1219: 1194: 1186: 1173: 1169: 1157: 1144: 1136: 1123: 1119: 1107: 1094: 1090: 1081: 1077: 1056: 1052: 1040: 1036: 1018: 997: 994: 993: 974: 971: 970: 967: 955:convex analysis 934: 931: 930: 914: 911: 910: 894: 891: 890: 853: 850: 849: 830: 827: 826: 801: 798: 797: 781: 778: 777: 754: 750: 748: 745: 744: 728: 725: 724: 704: 700: 698: 695: 694: 678: 675: 674: 658: 655: 654: 621: 617: 615: 612: 611: 583: 580: 579: 553: 549: 547: 544: 543: 523: 519: 510: 506: 485: 481: 469: 465: 447: 426: 423: 422: 399: 396: 395: 375: 372: 371: 364:Euclidean space 344: 341: 340: 321: 318: 317: 300: 297: 296: 266: 263: 262: 245: 242: 241: 225: 222: 221: 204: 201: 200: 174: 171: 170: 140: 137: 136: 121: 118: 117: 100: 97: 96: 79: 76: 75: 57: 54: 53: 36: 33: 32: 17: 12: 11: 5: 2754: 2744: 2743: 2738: 2733: 2728: 2712: 2711: 2700:(3): 197–229, 2684: 2666: 2652:10.1.1.28.9372 2645:(3): 435–450, 2629: 2609: 2590: 2569:Kruskal, J. B. 2559: 2558: 2556: 2553: 2540: 2537: 2534: 2514: 2492: 2489: 2484: 2481: 2477: 2454: 2451: 2447: 2425: 2403: 2380: 2377: 2373: 2356: 2353: 2349: 2348: 2336: 2333: 2330: 2325: 2321: 2317: 2314: 2311: 2308: 2303: 2300: 2297: 2293: 2289: 2286: 2275: 2264: 2261: 2258: 2255: 2252: 2249: 2244: 2240: 2236: 2231: 2227: 2216: 2203: 2200: 2197: 2193: 2189: 2186: 2164: 2161: 2157: 2140: 2139: 2128: 2125: 2122: 2119: 2116: 2113: 2110: 2107: 2104: 2101: 2098: 2095: 2091: 2088: 2083: 2079: 2076: 2073: 2070: 2066: 2063: 2058: 2054: 2051: 2048: 2038: 2027: 2024: 2021: 2018: 2015: 2011: 2008: 2003: 1999: 1996: 1993: 1990: 1986: 1983: 1978: 1974: 1971: 1968: 1965: 1962: 1959: 1956: 1933: 1930: 1927: 1924: 1921: 1918: 1904:Cauchy-Schwarz 1886: 1883: 1879: 1873: 1868: 1865: 1862: 1859: 1856: 1853: 1850: 1846: 1842: 1839: 1834: 1831: 1827: 1804: 1801: 1798: 1795: 1792: 1789: 1786: 1783: 1780: 1775: 1772: 1768: 1747: 1744: 1739: 1736: 1732: 1720: 1719: 1708: 1705: 1702: 1699: 1696: 1693: 1690: 1687: 1684: 1679: 1676: 1672: 1648: 1645: 1642: 1637: 1634: 1630: 1622: 1619: 1615: 1609: 1606: 1602: 1595: 1592: 1587: 1584: 1580: 1556: 1553: 1550: 1547: 1536: 1535: 1524: 1521: 1518: 1515: 1512: 1508: 1505: 1500: 1496: 1493: 1490: 1487: 1484: 1481: 1477: 1474: 1469: 1465: 1462: 1459: 1456: 1451: 1448: 1444: 1438: 1435: 1431: 1425: 1422: 1418: 1412: 1409: 1406: 1402: 1378: 1375: 1371: 1368: 1344: 1334:Hessian matrix 1332:(i.e. for the 1321: 1301: 1290: 1289: 1278: 1275: 1272: 1267: 1264: 1260: 1254: 1251: 1247: 1241: 1238: 1234: 1228: 1225: 1222: 1218: 1214: 1211: 1208: 1205: 1202: 1197: 1192: 1189: 1185: 1179: 1176: 1172: 1166: 1163: 1160: 1156: 1152: 1147: 1142: 1139: 1135: 1129: 1126: 1122: 1116: 1113: 1110: 1106: 1102: 1097: 1093: 1087: 1084: 1080: 1076: 1073: 1070: 1067: 1062: 1059: 1055: 1051: 1046: 1043: 1039: 1033: 1030: 1027: 1024: 1021: 1017: 1013: 1010: 1007: 1004: 1001: 978: 966: 963: 938: 918: 898: 885:. De Leeuw's 866: 863: 860: 857: 834: 814: 811: 808: 805: 785: 760: 757: 753: 732: 710: 707: 703: 682: 662: 638: 635: 632: 627: 624: 620: 599: 596: 593: 590: 587: 567: 564: 559: 556: 552: 540: 539: 526: 522: 516: 513: 509: 505: 502: 499: 496: 491: 488: 484: 480: 475: 472: 468: 462: 459: 456: 453: 450: 446: 442: 439: 436: 433: 430: 403: 379: 351: 348: 328: 325: 304: 282: 279: 276: 273: 270: 249: 229: 208: 187: 184: 181: 178: 153: 150: 147: 144: 125: 104: 83: 61: 40: 15: 9: 6: 4: 3: 2: 2753: 2742: 2739: 2737: 2734: 2732: 2729: 2727: 2726:Graph drawing 2724: 2723: 2721: 2707: 2703: 2699: 2695: 2688: 2679: 2678: 2670: 2662: 2658: 2653: 2648: 2644: 2640: 2633: 2624: 2620: 2613: 2604: 2597: 2595: 2586: 2582: 2578: 2574: 2573:Psychometrika 2570: 2564: 2560: 2552: 2538: 2535: 2532: 2512: 2490: 2487: 2482: 2479: 2475: 2452: 2449: 2445: 2436: 2423: 2414: 2401: 2378: 2375: 2371: 2362: 2361:graph drawing 2352: 2334: 2331: 2323: 2319: 2312: 2309: 2301: 2298: 2295: 2291: 2284: 2276: 2259: 2256: 2253: 2247: 2242: 2229: 2225: 2217: 2201: 2198: 2195: 2191: 2184: 2162: 2159: 2155: 2146: 2145: 2144: 2123: 2120: 2117: 2111: 2108: 2105: 2099: 2093: 2089: 2086: 2081: 2077: 2074: 2071: 2068: 2064: 2061: 2056: 2052: 2049: 2046: 2039: 2025: 2019: 2013: 2009: 2006: 2001: 1997: 1994: 1991: 1988: 1984: 1981: 1976: 1972: 1969: 1966: 1960: 1954: 1947: 1946: 1945: 1928: 1925: 1922: 1916: 1907: 1905: 1900: 1884: 1881: 1877: 1871: 1866: 1863: 1860: 1857: 1854: 1851: 1848: 1844: 1840: 1837: 1832: 1829: 1825: 1815: 1802: 1799: 1796: 1793: 1790: 1787: 1781: 1773: 1770: 1766: 1745: 1742: 1737: 1734: 1730: 1706: 1703: 1700: 1697: 1694: 1691: 1685: 1677: 1674: 1670: 1643: 1635: 1632: 1628: 1620: 1617: 1613: 1607: 1604: 1600: 1593: 1590: 1585: 1582: 1578: 1570: 1569: 1568: 1551: 1545: 1522: 1516: 1510: 1506: 1503: 1498: 1494: 1491: 1485: 1479: 1475: 1472: 1467: 1463: 1457: 1449: 1446: 1442: 1436: 1433: 1429: 1423: 1420: 1416: 1410: 1407: 1404: 1400: 1392: 1391: 1390: 1376: 1373: 1369: 1366: 1358: 1342: 1335: 1319: 1299: 1273: 1265: 1262: 1258: 1252: 1249: 1245: 1239: 1236: 1232: 1226: 1223: 1220: 1216: 1212: 1209: 1203: 1195: 1190: 1187: 1183: 1177: 1174: 1170: 1164: 1161: 1158: 1154: 1150: 1145: 1140: 1137: 1133: 1127: 1124: 1120: 1114: 1111: 1108: 1104: 1100: 1095: 1085: 1082: 1078: 1074: 1068: 1060: 1057: 1053: 1044: 1041: 1037: 1031: 1028: 1025: 1022: 1019: 1015: 1011: 1005: 999: 992: 991: 990: 976: 962: 960: 956: 952: 949:, called the 936: 916: 896: 888: 884: 880: 861: 855: 846: 832: 809: 803: 783: 774: 758: 755: 751: 730: 708: 705: 701: 680: 660: 652: 633: 625: 622: 618: 594: 591: 588: 565: 562: 557: 554: 550: 524: 514: 511: 507: 503: 497: 489: 486: 482: 473: 470: 466: 460: 457: 454: 451: 448: 444: 440: 434: 428: 421: 420: 419: 417: 401: 393: 392:loss function 390:is a cost or 377: 369: 365: 349: 346: 326: 323: 315: 302: 293: 277: 274: 271: 247: 227: 219: 206: 182: 176: 168: 164: 148: 145: 123: 102: 94: 81: 72: 59: 51: 38: 29: 25: 21: 2697: 2693: 2687: 2676: 2669: 2642: 2638: 2632: 2622: 2612: 2602: 2576: 2572: 2563: 2416: 2394: 2358: 2350: 2177:step we set 2141: 1908: 1901: 1816: 1721: 1537: 1291: 968: 958: 950: 886: 883:Jan de Leeuw 847: 775: 541: 415: 362:dimensional 295: 261: 199: 166: 116: 74: 52: 31: 19: 18: 2619:Groenen, P. 2579:(1): 1–27, 929:at a point 260:, i.e. the 198:. Usually 2720:Categories 2617:Borg, I.; 2555:References 959:majorizing 115:points in 2647:CiteSeerX 2533:α 2513:α 2505:. Here, 2491:α 2488:− 2476:δ 2372:δ 2335:ϵ 2313:σ 2310:− 2299:− 2285:σ 2248:τ 2235:← 2199:− 2188:← 2112:τ 2075:− 2047:≤ 1995:− 1955:σ 1917:τ 1864:≠ 1845:∑ 1841:− 1800:≠ 1704:≠ 1692:≠ 1614:δ 1594:− 1495:≥ 1430:δ 1401:∑ 1246:δ 1217:∑ 1210:− 1155:∑ 1134:δ 1105:∑ 1079:δ 1075:− 1029:≤ 1016:∑ 1000:σ 977:σ 917:σ 897:σ 856:σ 804:σ 702:δ 563:≥ 508:δ 504:− 458:≤ 445:∑ 429:σ 378:σ 350:− 327:− 275:× 177:σ 169:function 146:≪ 2621:(1997), 2277:stop if 2090:′ 2065:′ 2010:′ 1985:′ 1507:′ 1476:′ 1370:′ 653:between 368:MDS plot 26:used in 2147:at the 649:is the 294:matrix 2649:  1538:where 953:. In 542:where 167:stress 22:is an 1567:has: 2415:and 2332:< 1817:and 1758:for 1722:and 1662:for 1408:< 1224:< 1162:< 1112:< 1023:< 693:and 673:and 452:< 2702:doi 2657:doi 2581:doi 2239:min 339:or 240:or 220:is 95:of 2722:: 2696:, 2655:, 2643:16 2641:, 2593:^ 2577:29 2575:, 2551:. 2082:tr 2057:tr 2002:tr 1977:tr 1899:. 1499:tr 1468:tr 1357:tr 610:, 2709:. 2704:: 2698:4 2682:. 2664:. 2659:: 2627:. 2607:. 2588:. 2583:: 2539:2 2536:= 2483:j 2480:i 2453:j 2450:i 2446:w 2424:j 2402:i 2379:j 2376:i 2329:) 2324:k 2320:X 2316:( 2307:) 2302:1 2296:k 2292:X 2288:( 2263:) 2260:Z 2257:, 2254:X 2251:( 2243:X 2230:k 2226:X 2202:1 2196:k 2192:X 2185:Z 2163:h 2160:t 2156:k 2127:) 2124:Z 2121:, 2118:X 2115:( 2109:= 2106:Z 2103:) 2100:Z 2097:( 2094:B 2087:X 2078:2 2072:X 2069:V 2062:X 2053:+ 2050:C 2026:X 2023:) 2020:X 2017:( 2014:B 2007:X 1998:2 1992:X 1989:V 1982:X 1973:+ 1970:C 1967:= 1964:) 1961:X 1958:( 1932:) 1929:Z 1926:, 1923:X 1920:( 1885:j 1882:i 1878:b 1872:n 1867:i 1861:j 1858:, 1855:1 1852:= 1849:j 1838:= 1833:i 1830:i 1826:b 1803:j 1797:i 1794:, 1791:0 1788:= 1785:) 1782:Z 1779:( 1774:j 1771:i 1767:d 1746:0 1743:= 1738:j 1735:i 1731:b 1707:j 1701:i 1698:, 1695:0 1689:) 1686:Z 1683:( 1678:j 1675:i 1671:d 1647:) 1644:Z 1641:( 1636:j 1633:i 1629:d 1621:j 1618:i 1608:j 1605:i 1601:w 1591:= 1586:j 1583:i 1579:b 1555:) 1552:Z 1549:( 1546:B 1523:Z 1520:) 1517:Z 1514:( 1511:B 1504:X 1492:X 1489:) 1486:X 1483:( 1480:B 1473:X 1464:= 1461:) 1458:X 1455:( 1450:j 1447:i 1443:d 1437:j 1434:i 1424:j 1421:i 1417:w 1411:j 1405:i 1377:X 1374:V 1367:X 1343:V 1320:X 1300:C 1277:) 1274:X 1271:( 1266:j 1263:i 1259:d 1253:j 1250:i 1240:j 1237:i 1233:w 1227:j 1221:i 1213:2 1207:) 1204:X 1201:( 1196:2 1191:j 1188:i 1184:d 1178:j 1175:i 1171:w 1165:j 1159:i 1151:+ 1146:2 1141:j 1138:i 1128:j 1125:i 1121:w 1115:j 1109:i 1101:= 1096:2 1092:) 1086:j 1083:i 1072:) 1069:X 1066:( 1061:j 1058:i 1054:d 1050:( 1045:j 1042:i 1038:w 1032:n 1026:j 1020:i 1012:= 1009:) 1006:X 1003:( 937:Z 865:) 862:X 859:( 833:m 813:) 810:X 807:( 784:X 759:j 756:i 752:w 731:m 709:j 706:i 681:j 661:i 637:) 634:X 631:( 626:j 623:i 619:d 598:) 595:j 592:, 589:i 586:( 566:0 558:j 555:i 551:w 525:2 521:) 515:j 512:i 501:) 498:X 495:( 490:j 487:i 483:d 479:( 474:j 471:i 467:w 461:n 455:j 449:i 441:= 438:) 435:X 432:( 416:r 402:m 347:3 324:2 303:X 281:) 278:r 272:n 269:( 248:3 228:2 207:r 186:) 183:X 180:( 152:) 149:m 143:( 124:r 103:n 82:X 60:m 39:n

Index

optimization strategy
multidimensional scaling
Euclidean space
MDS plot
loss function
euclidean distance
steepest descent
Jan de Leeuw
convex analysis
Hessian matrix
tr
Cauchy-Schwarz
graph drawing
Kruskal, J. B.
doi
10.1007/BF02289565


Groenen, P.
CiteSeerX
10.1.1.28.9372
doi
10.1007/s001800100077
Proceedings of 12th Int. Symp. Graph Drawing (GD'04)
doi
10.1145/264645.264657
Categories
Graph drawing
Dimension reduction
Mathematical optimization

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