### Table 2: Reduced-order control nc Worst-case Nominal

"... In PAGE 24: ... Note that the case with a zero-order lter corresponds to the robust H2 problem studied by Stoorvogel [17], where the uncertainty set is assumed to include nonlinear and possibly noncausal perturbations. Table2 shows the performance achieved with reduced-order controllers of various orders nc. The uncertainty radius had the value = 0:5.... ..."

### Table 1: Nominal Speed-Up for N-gram Models nominal speed-up for training the language model (1) with non-nested features for the Switchboard. The computa- tional complexity is also nominal to reduce in 2 orders of magnitude using the new algorithm. IIS New Speed-up

2000

Cited by 2

### Table 1: Nominal Speed-Up for N-gram Models nominal speed-up for training the language model (1) with non-nested features for the Switchboard. The computa- tional complexity is also nominal to reduce in 2 orders of magnitude using the new algorithm. IIS New Speed-up

2000

Cited by 2

### Table 10: A simple (nominal-order)-order information table

"... In PAGE 21: ...Table10... ..."

### Table 4: Seventh Order (P; )-admissible Controllers Hence the nominal 20 state (P; )-admissible controller has successfully been reduced down to a 7 state (P; )-admissible controller. Controller complexity has therefore been (dramatically) reduced without an associated loss of performance. Figure 9 gives a plot of for the nominal controller K0 and for the seventh order controller K. As designed, both controllers give lt; 1 and hence perform robustly (for all G 2 RH1, k Gk1 lt; 1) with the unscaled plant. A comparison of frequency domain characteristics and time domain responses for the full and reduced order control laws is given in [16]. They show that no signi cant degradation is introduced into any of the performance measures. 3Here we were required to switch between optimising over the B and the D matrices and the C and the D matrices of K. Both of these optimisations are convex.

### Table 3. Reduced bias in Pfa by using the multilook correction factor, k.

"... In PAGE 3: ...able 2. Reduced bias in Pfa by introducing a correction factor, k. distributed speckle and multi-pixel targets is somewhat more complicated and an outline of the derivation is given in Appendix B. However, we can verify the correctness of using this approach from Table3 which shows... In PAGE 4: ...able 6. False alarm rates for segann for di erent order speckle and target sizes. Table 6 allows us to compare the false alarm rates for segann with CFAR (cf. Table3 ) for a nominal Pd... ..."

### Table 9: Number of oating point operations, compensator computations. Full Order Reduced Comp. Reduced Order N

2000

"... In PAGE 25: ...same storage (if M=P as we have done here) and the same number of ops. In Table9 , we show the number of ops to compute the full order compensator, the number to compute the reduced order compensator including the grid selection algorithm, and the number required to compute the reduced order compensator if the grid is already known. In practice, the grid selection algorithm would be performed once.... ..."

Cited by 7

### Table 4.1 Convergence of n = 2 Problems 1st entry: No. of f evaluations to reduce 11 f /I2 below 2nd entry: Max. absolute component error in final estimate of inverse Jacobian at solution.

1971

### Table 4.4 Convergence of n = 15 Problems 1st entry: No. apos;of f evaluations to reduce 11 f 11, below 2nd entry: Max. absolute component error in final estimate of inverse Jacobian at solution. .. .. . - -.. . ..

1971

### TABLE III: CONVENTIONAL BASELINE HSCT DATUM POINTS

1999

Cited by 8