"... In PAGE 5: ... When the flash fills up or a read- miss to it occurs, the corresponding disk is spun-up and the flash sectors are flushed to their respective locations on disk. We also built a a simulator to model a hybrid disk (power state specifications in Table1 ) to allow expedi- ent evaluation of the proposed enhancements with sev- eral week-long block-level traces. The simulator consid- ers power relative to the given trace for different power states: read/write, seek, idle, standby, and spin-up.... ..."

"... In PAGE 6: ... Similarly, Boreus provides a highly structured environment because of the deterministic object trajec- tory and his fully predictive gaze direction. Columns \N quot; (normal) of Table1 summarize the simulation re- sults for this experiment for both infants. As one can see, the results are very similar for Optimiss.... In PAGE 7: ... In this paper we used the following reward structures: Autist: Rfrontal = 1, Rpro le = 0; Williams syndrome: Rfrontal = 2, Rpro le = 2. Columns \A quot;(autist) and \W quot;(Williams syndrome) of Table1 summarize the simulation results for this ex- periment for both infants. In addition, Figure 5 illus- trates the GFI of Optimuse as an autist and as a toddler with Williams syndrome.... In PAGE 7: ...utual gaze contact (i.e., avoids negative rewards for the infant), whereas Careus succeeds in \guiding away quot; the Williams toddler from staring at his face to the object. From Table1 we can also see that the random care- giver Randomus is unsuccessful in all situations. Look- ing at Optimuse, we nd that Boreus performs even better than Avoidus for the autist because he too never looks at the infant, but provides an even more deter- ministic behavior.... ..."

"... In PAGE 18: ... 6 Experiments We rst show a comparison in di erent practical experiments among those methods which deal with nested loops containing conditional constructs. The experiments presented in Table1 are the Floyd- Steinberg algorithm, the MD CdDFG from [29], two MD CdDFGs from [12], the three-value median lter from [31]. These algorithms come from diverse areas and are typical for Multi-Dimensional applications.... In PAGE 18: ... The methods listed are list scheduling, GPMB approach from [31], Multi-Dimensional Branch Predication [5, 28, 32, 33] which utilizes the OPTIMUS algorithm from [22], polynomial time Branch Anticipation algorithm MDBA, and the optimal schedule length utilizing resource sharing but ignoring any sharing-prevention cycles. In Table1 , the nodes column lists the total number of operations in each MD CdDFG. The len column under each method gives the schedule length for the respective MD CdDFG.... ..."