"... In PAGE 20: ... In each of these data sets we run the peeling procedure using the maximum surprise score of Sec- tion 4 with the TNoM score. Table1 summarizes the scores of the top discovered classifications using the various scoring mechanisms we discussed above and also summarize their difference to the published classification of the data sets. Note that LOOCV evaluation in this table is not an in- dependent statistical validation of the discovered partitions, since information about the expression profile of any sample is effecting the classifier learned from any m ; 1 set.... In PAGE 20: ... On the leukemia data set we run our search procedure with the additional constraint that it should only examine labeling without control tissues. Peeling found six labelings, the first four of which are shown in Table1 . All six labelings score better than the original labeling by the max- surprise score, and by the number of significant genes.... In PAGE 21: ... (Clinical risk is evaluated using international prognostic index, a standard medical index, evaluated at the time the sample was taken.) In Figure 4 we plot survival rates for patients for the four putative DLBCL classifications described in Table1 . As we can see, some of the classifications, such as the forth one, are not predictive about patient survival.... In PAGE 26: ...8 0.9 1 19 patients 9 deaths 5 patients 2 deaths Figure 4: Kaplan-Meier survival plots for the 4 DLBCL classifications described in Table1 . The x-axis is the number of years after the samples were taken, and the y-axis is the fraction of patients survived so far.... ..."

"... In PAGE 20: ... In each of these data sets we run the peeling procedure using the maximum surprise score of Sec- tion 4 with the TNoM score. Table1 summarizes the scores of the top discovered classifications using the various scoring mechanisms we discussed above and also summarize their difference to the published classification of the data sets. Note that LOOCV evaluation in this table is not an in- dependent statistical validation of the discovered partitions, since information about the expression profile of any sample is effecting the classifier learned from any m ? 1 set.... In PAGE 20: ... On the leukemia data set we run our search procedure with the additional constraint that it should only examine labeling without control tissues. Peeling found six labelings, the first four of which are shown in Table1 . All six labelings score better than the original labeling by the max- surprise score, and by the number of significant genes.... In PAGE 21: ... (Clinical risk is evaluated using international prognostic index, a standard medical index, evaluated at the time the sample was taken.) In Figure 4 we plot survival rates for patients for the four putative DLBCL classifications described in Table1 . As we can see, some of the classifications, such as the forth one, are not predictive about patient survival.... In PAGE 26: ...8 0.9 1 19 patients 9 deaths 5 patients 2 deaths Figure 4: Kaplan-Meier survival plots for the 4 DLBCL classifications described in Table1 . The x-axis is the number of years after the samples were taken, and the y-axis is the fraction of patients survived so far.... ..."

"... In PAGE 11: ....0001 for the largest nuclear area and was 0.0092 for the number of metastatic axillary lymph nodes. An analysis of deviance 16 was performed in order to estimate the importance and interdependence of nuclear features relative to lymph node status ( Table2 ). For a single variable model, Worst Area is a better prognosticator than is lymph node status (Table 2, rows 1 and 2).... In PAGE 11: ... An analysis of deviance 16 was performed in order to estimate the importance and interdependence of nuclear features relative to lymph node status (Table 2). For a single variable model, Worst Area is a better prognosticator than is lymph node status ( Table2 , rows 1 and 2). ... In PAGE 12: ...given in Table2 , row 3. Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5).... In PAGE 12: ...given in Table 2, row 3. Adjusting for Worst Area causes a large effect ( Table2 , row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5). The adjusted numbers (Table 2, rows 4 and 5) are about the same as are the unadjusted ones (Table 2, rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another.... In PAGE 12: ...given in Table 2, row 3. Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less ( Table2 , row 5). The adjusted numbers (Table 2, rows 4 and 5) are about the same as are the unadjusted ones (Table 2, rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another.... In PAGE 12: ... Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5). The adjusted numbers ( Table2 , rows 4 and 5) are about the same as are the unadjusted ones (Table 2, rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another. When machine learning with leave-one-out cross validation was used to select the optimal number of features for prognostic models, computer-derived nuclear Worst Radius was used 84.... In PAGE 12: ... Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5). The adjusted numbers (Table 2, rows 4 and 5) are about the same as are the unadjusted ones ( Table2 , rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another. When machine learning with leave-one-out cross validation was used to select the optimal number of features for prognostic models, computer-derived nuclear Worst Radius was used 84.... ..."

"... In PAGE 11: ....0001 for the largest nuclear area and was 0.0092 for the number of metastatic axillary lymph nodes. An analysis of deviance 16 was performed in order to estimate the importance and interdependence of nuclear features relative to lymph node status ( Table2 ). For a single variable model, Worst Area is a better prognosticator than is lymph node status (Table 2, rows 1 and 2).... In PAGE 11: ... An analysis of deviance 16 was performed in order to estimate the importance and interdependence of nuclear features relative to lymph node status (Table 2). For a single variable model, Worst Area is a better prognosticator than is lymph node status ( Table2 , rows 1 and 2). ... In PAGE 12: ...given in Table2 , row 3. Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5).... In PAGE 12: ...given in Table 2, row 3. Adjusting for Worst Area causes a large effect ( Table2 , row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5). The adjusted numbers (Table 2, rows 4 and 5) are about the same as are the unadjusted ones (Table 2, rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another.... In PAGE 12: ...given in Table 2, row 3. Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less ( Table2 , row 5). The adjusted numbers (Table 2, rows 4 and 5) are about the same as are the unadjusted ones (Table 2, rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another.... In PAGE 12: ... Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5). The adjusted numbers ( Table2 , rows 4 and 5) are about the same as are the unadjusted ones (Table 2, rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another. When machine learning with leave-one-out cross validation was used to select the optimal number of features for prognostic models, computer-derived nuclear Worst Radius was used 84.... In PAGE 12: ... Adjusting for Worst Area causes a large effect (Table 2, row 4), where as the effect of adjusting for lymph node status is less (Table 2, row 5). The adjusted numbers (Table 2, rows 4 and 5) are about the same as are the unadjusted ones ( Table2 , rows 1 and 2 respectively), indicating that the contributions by Worst Area and lymph node status are independent of one another. When machine learning with leave-one-out cross validation was used to select the optimal number of features for prognostic models, computer-derived nuclear Worst Radius was used 84.... ..."

"... In PAGE 87: ...A-19 Table2 summarizes the results from the third season of the methane addition experiments, and the experiments with phenol and toluene as primary substrates. As shown in the table, the use of phenol and toluene achieved higher percent removals of TCE (93 - 94%) compared with use of methane (19%).... In PAGE 92: ...67 - 1 44 17 - 117 T2 38 13.4 1 29 47 - 63 Table2 . Operating Parameters for Balanced Flow Operation (Days 317 - 444) [3] Treatment Well Groundwater Pumping Rate (L/min) Toluene Addition (mg/L) Toluene Addition - Pulses per Day DO Addition (mg/L) H2O2 Addition (mg/L) T1 25 9.... In PAGE 100: ... The system was operational 90% of the time and no problems were reported during the demonstration. Technology Cost [1,5,6] Table2 presents the projected costs for full-scale application of MEBR. The projected capital costs were $452,407 (including equipment costs amortized over 10 years, and costs for well installation and mobilization), and the projected operation and maintenance (O amp;M) costs were $236,465 (including monitoring, consumables, and demobilization).... In PAGE 100: ... The projected capital costs were $452,407 (including equipment costs amortized over 10 years, and costs for well installation and mobilization), and the projected operation and maintenance (O amp;M) costs were $236,465 (including monitoring, consumables, and demobilization). Table2 . Project Costs for Full-scale MEBR Application [1,5] Element Cost ($) Capital Site cost 5,400 Equipment cost 9,200 Design and Engineering 10,000 Mobile equipment 18,000 Well Installation 183,000 Other fixed equipment 183,732 Mobilization 43,075 Total Capital Equipment and Mobilization Cost 452,407 O amp;M Costs Monitoring/maintenance 71,175 Consumables 122,215 Demobilization 43,075 Total O amp;M 236,465 Summary Observations and Lessons Learned [1,2] The in situ bioremediation system demonstrated as SRS removed about 17,000 lbs of VOCs through vacuum extraction (about 12,000 lbs) and through biodegradation (about 5,000 lbs).... In PAGE 106: ... No specific cleanup goals have been identified for groundwater at this site. Table2 summarizes groundwater monitoring data from June 1995 (prior to system operation) to December 1998 for five sampling events, including data for TCE, DCE, and VC, as well as other parameters such as chloride, DO, and redox potential. The data presented are the average concentrations measured in six wells located outside of the treatment zone ( Outside Wells ) and in eight wells located within the treatment zone ( Inside Wells ), including one of the monitoring wells, well MW-40, that was located within the apparent source area.... In PAGE 107: ...A-39 Table2 . Summary of Technology Performance Data Average Concentrations (mg/L) [3] Parameter Analytical Methods* Jun-95 May-96 Jun-97 Mar-98 Dec-98 Inside Wells Outside Wells Inside Wells Outside Wells Inside Wells Outside Wells Inside Wells Outside Wells Inside Wells Outside Wells Average including potential source well MW-40 TCE EPA 8260 22.... In PAGE 108: ...A-40 The concentration of chloride was used to evaluate the potential impact of dilution on groundwater quality. As shown on Table2 , there is a difference between the chloride concentrations in the inside wells and those in the outside wells. According to the site contractor, assuming the chloride is due exclusively to dechlorination of TCE (and its degradation products), the average TCE concentration at project outset was estimated to be as high as 100 mg/L to 125 mg/L.... ..."