"... In PAGE 1: ... Results The results are presented in Figures and Tables in the following sections. The figures and the temporal data in Table 2 are based on linear data whereas Table1 is based on logarithmic... In PAGE 4: ... Table1 . Statistical analysis of the results with reference to temporal productions (see text).... In PAGE 4: ... Discussion Language, gender and focus tonal productions were found significant for (utterance) tonal onset and word peak f0 but hardly for (utterance) tonal offset (gender productions are however significant). Significant interactions were also found, most importantly between language and speaker gender as well as language and focus with reference to word peak f0 (see Table1 ). Temporal productions were also found significant as a function of language, gender and focus whereas significant interactions were mostly evident with reference to vowel segment productions (see Table 2).... ..."

"... In PAGE 4: ...2. Species that are invasive elsewhere and are likewise invasive or have the potential to become so in American Samoa Some known invasive plants that are causing trouble in similar ecosystems have been introduced into American Samoa (see Appendix 2, Table2 ). Some of them are already causing problems while others are not.... ..."

"... In PAGE 51: ... Each earth model is approximately 1000 feet long and consists of 80 beds. There is a total of 19 models for training, testing and validation as shown in Table5 . During training one calculates the error on a set of examples that is not used during training, the so-called validation set.... In PAGE 74: ...2 % 77 % correct 47485 Shallow-log (2): After this training we add models without invasion and some models with invasion and restart the training. For this purpose we use the whole training set (see Table5 ) (the target logs are now sampled every 1.0 feet) and 40 models without invasion (target logs wer sampled every 0.... ..."

"... In PAGE 8: ... If ma21 xa8 aa22 has more than one element, this set will give us the possibilities for a value aa21 xa22 , based on previous experience. The steps taken by this method for the example data are given in Table4 . We have listed the interme- diate sets ma21 xa8 aa22 to indicate how the compatibilities change after we have extended attribute functions in one step.... ..."

"... In PAGE 6: ... If ma21 xa8 aa22 has more than one element, this set will give us the possibilities for a value aa21 xa22 , based on previous experience. The steps taken by this method for the example data are given in Table4 . We have listed the interme- diate sets ma21 xa8 aa22 to indicate how the compatibilities change after we have extended attribute functions in one step.... ..."

"... In PAGE 5: ...able 3: Listed Plant Species....................................................................................................... 15 Table4 : Invasive Exotic Plant Species and Treatment Recomendations .... ..."

"... In PAGE 5: ... Invasion was quantified as the percentage of bacteria inoculated into the tissue culture wells which were recovered from lysed Henle cells. As shown in Table1 , a control Salmonella strain, SB136, with a null mutation in invA and a previously documented functional defect in type III secretion and host cell invasion (4), achieves lt;1% invasion into Henle cells, independent of arabinose addition. Salmonella SB300A#1 transformants which express an EGS complementary to mRNA encoding synthetic C5 (also shown as a negative control for type III secretion assays, above) have a gt;20% rate of invasion, with a small decrease in invasion observed after arabinose addition.... ..."

"... In PAGE 72: ...Experimenting with architecture constraints 63 Table2 0: Results for convolutional- regression nets (1). Convolutional 1 Test 20 Test 21 architecture 3 60:6 9:5:1 3 150:6 27:15:1 number of weights 503 2827 sampling period 0.... In PAGE 72: ...7 % 25.5 % Table2 1: Results for convolutional- regression nets (2). Convolutional 2 Test 22 Test 23 architecture 3 100:6 16:15:1 3 150:9 27:15:1 number of weights 1927 3865 sampling period 0.... In PAGE 74: ...re sampled every 0.2 feet. The results are given in Table 22. Table2 2: Average relative error and performance for earth models with invasion (Shallow-log). average percentage of set number of relative error predicted correct patterns Training set 4.... In PAGE 75: ... FORWARD MODELLING WITH MUD INVASION 45 50 55 60 65 70 75 80 85 90 95 100 training validation testing percentage of log that is correct Non-Uniform Haar Uniform Convolutional-regression Figure 40: Intermediate results (2) of training di erent neural nets on 6 models. Table2 3: Average relative error and performance for earth models with and without invasion (Shallow-log). average percentage of set number of relative error predicted correct patterns invasion Training set 5.... In PAGE 76: ... The results are given in Table 24. Table2 4: Average relative error and performance for earth models with and without invasion (Deep-log). average percentage of set number of relative error predicted correct patterns invasion Training set 8.... ..."

"... In PAGE 9: ... We tested whether CSR1 expression affects prostate cancer cell invasiveness in an in vitro Matrigel transmi- gration experiment and in an in vivo mouse tumorigenesis experiment. Three-fold fewer CSR1-expressing PC3 cells (pCMV-CSR1 transformants) traversed the Matrigel than did control cells ( Table2 ). Moreover, the propensity for PC3 cells injected into mice to develop into tumors was suppressed by CSR1 expression (Table 3).... ..."