"... In PAGE 3: ...Figure 1 provides a graph of the historic returns and risk for different asset classes from 1970 to 1994. Table1 presents the data for each asset class. Figure 1 plots such assets as small-cap stocks, mid-cap stocks, S amp;P 500 stocks, gold, intermediate corporate and government bonds, long term government bonds and 90 day treasury bills.... ..."

"... In PAGE 2: ... As part of this evolution, the office itself is finally undergoing a long overdue re-conceptualisation. Table1 captures the evolution in office space and ways of working. From this summary a variety of physical layouts are implied.... ..."

"... In PAGE 2: ... Traditional TCP connec- tions are unable to achieve high throughput in highspeed wide area networks due to the long packet loss recovery times and the need for low supporting loss rates. Table1 shows the properties of a traditional TCP connection with a round trip time of 200ms and a segment size of 1500 bytes. A packet loss rate of 10 7 is compa- rable with those that can occur on long haul fiber links, within net- work devices, and in end-systems; this places a limit on throughput before any transient congestion due to load fluctuations are con- sidered.... ..."

"... In PAGE 14: ...The regional breakdown of capital inflows to developing countries reveals that the surge phenomenon is widespread, but is especially pronounced in East Asia and Latin America. To show this, in Table2 below we allocate the long-term private net flows reported in Table I into the regions traditionally analyzed by the World Bank. Table 2 Annual Long-Term Private Capital Net flows (By Region)... In PAGE 14: ... In both cases, the pace of inflows accelerated after 1991. There is a suggestion in Table2 that the phenomenon may recently have become more widespread, reaching South Asia as well as Sub-Saharan Africa in 1993.12 12 Impressionistic evidence suggests that the phenomenon has recently become important in India and Pakistan, as well as in Kenya and Uganda.... ..."

"... In PAGE 3: ... Comparisons against reinforcement learn- ing, A*, and PEGASUS are summarized in Table 2 and dis- cussed in more detail later (while we do not take an RL ap- proach in our technique, we believe it is valuable to compare against RL since it such a popular approach). The results of an informal user case study are reported in Table3 . Videos of the animation results are available from: http://rivit.... ..."

"... In PAGE 3: ... Comparisons against reinforcement learn- ing, A*, and PEGASUS are summarized in Table 2 and dis- cussed in more detail later (while we do not take an RL ap- proach in our technique, we believe it is valuable to compare against RL since it such a popular approach). The results of an informal user case study are reported in Table3 . Videos of the animation results are available from: http://rivit.... ..."

"... In PAGE 3: ... A possible way to find such keywords or phrases is to find the keywords that frequently appear in the target set as in traditional data mining. However, this does not works in most text collections because in a typical English text, the most frequent keywords are stopwords like the or an (see Table1 (a) and Table 3 (a)). These keywords are basic constituents of English grammars and convey no information on the contents of the text collection.... In PAGE 3: ... We can easily observe that most stopwords appear evenly in the target and the control set, while informative keywords appear more frequently in the target set than the control set. Therefore, the optimized pattern discovery algorithm will find those keywords or phrases that appear more frequently in the target set than the control set by minimizing a given statistical measure such as the information entropy or the prediction error (See Table1 (b)-(c) and Table 3 (b)-... In PAGE 6: ...2.2 The first experiment In Table1 , we show the list of the phrase patterns discovered by our mining system, which capture the category ship relative to other categories of Reuters newswires. In Fig.... In PAGE 6: ...ontain the topic keywords in the major news stories for the period in 1987 (Fig. 1(b)). Such keywords are hard to find by traditional frequent pattern discovery because of the existence of the high frequency words such as CWtheCX and CWareCX. The patterns of medium rank (BEBIBD AO BEBJBC) are long phrases, such that CWlloyds shipping intelligenceCX and CWiranian oil platformCX, as a summary ( Table1 (c)), which cannot be represented by any combination of non-contiguous keywords. 3.... ..."

"... In PAGE 2: ... A possible way to find such keywords or phrases is to find the keywords that frequently appear in the target set as in traditional data mining. However, this does not works in most text collections because in a typical English text, the most frequent keywords are stopwords like the or an (see Table1 (a) and Table 3 (a)). These keywords are basic constituents of English grammars and convey no informa- tion on the contents of the text collection.... In PAGE 2: ... We can easily observe that most stopwords appear evenly in the target and the con- trol set, while informative keywords appear more frequently in the target set than the control set. Therefore, the opti- mized pattern discovery algorithm will find those keywords or phrases that appear more frequently in the target set than the control set by minimizing a given statistical measure such as the information entropy or the prediction error (See Table1 (b)-(c) and Table 3 (b)-(c)). 2.... In PAGE 4: ...2. The first experiment In Table1 , we show the list of the phrase patterns dis- covered by our mining system, which capture the category ship relative to other categories of Reuters newswires. In Fig.... ..."

"... In PAGE 21: ...0091 long-chain fatty acid biosynthesis Myo5a, Plp1 8 0048066 P .0091 pigmentation during development Gnaq, Myo5a Table2 : Clusters discovered with a traditional mixture model (independent clustering). Type corresponds to either (P)rocess or (F)unction.... ..."

"... In PAGE 3: ... Table2 shows the results presented in [13], where the latency experienced by a lightweight RPC (Remote proce- dure Call) protocol is analyzed with Ethernet, FDDI and ATM, as before. In particular, the different contributions to latency are separately studied, and two sizes of messages have been considered, which are labeled short and long .... In PAGE 3: ... In particular, the different contributions to latency are separately studied, and two sizes of messages have been considered, which are labeled short and long . The speed improvements achieved by each contribution to latency with respect to the performance over Ethernet have been included in Table2 . These results show that the 14- /19-foldbandwidthincrement of ATM results in an effective speed increase of a factor of only 2, highlighting the influ- ence of both hardware and software concerns in addition to the networkingtechnologyitself.... ..."