Discussion of the results. Critical design change decisions made by MTI after the baseline studies and first version of the DIP, and again after the MTE, significantly contributed to the successful increases in the behavior change and health system improvements achieved by the project. Addition of a malaria intervention was an essential response to address the epidemiologic patterns revealed in the baseline assessments as well as feedback from the DIP reviewers. A staffing change was made when MTI assumed responsibility for hiring and mentoring the Community Outreach Coordinator that was originally under XXXX’s supervision. IR 1: Communities assume responsibility for their own health with strengthened community structures, linkages with Health Facility staff, and enhanced human resource capacity Intermediate Result 1: Communities assume responsibility for their own health with strengthened community structures, linkages with Health Facility staff, and enhanced human resource capacity Six of the seven project indicators for strengthening community capacity were met. The community referral system was working well and CHCs had been revitalized and implementing emergency transport systems and economic plans. The indicator “% of health facilities with active CHDCs who have met in the last three months” was not met as CHDCs were established or revitalized at only 23 of 30 health facilities. MTI offered assistance to the CHT to establish CHCDs at the remaining seven facilities, though the CHT has not taken the initiative to do so. The DIP process that led to adopting the Care Group approach after the beginning of the program is a lesson in sustainable capacity building. XXX contracted as a consultant and trainer an experienced Liberian nurse midwife who was familiar with community-based MCH programs and Care Groups and who also had years of experience working with both the MoHSW and NGO programs. The project was able to introduce Care Groups in GCMC in a culturally- sensitive and practical way. The approach makes it possible to reach every woman in the target population on a regular basis without overwhelming community volunteers and has proven feasible and effective in increasing child survival intervention coverage in case after case in multiple community contexts in Africa and beyond. Although not yet thoroughly studied, there are anecdotal reports from the MTI CSP, and other similarly designed projects being involved as a Care Group volunteer enhances women volunteers’ self-esteem, and...
Discussion of the results. First, the reader should notice that the free energy of Sparse MBP (24) and of Xxxxxx QAP (25) exhibit a phase transition similar to that of Xxxxxxx’x Random Energy Model (REM) (Xxxxxxx, 1981, Section V). We like to emphasize that Sparse MBP and Xxxxxx QAP introduce some correlation between costs of pairs of solutions, while REM defines a technically much simpler setting without any correlations between cost values. Second, regarding the behavior of expected log-posterior agreement (4) shown in Theorem 4, we can notice that it shows two phase transitions with quadratic, mixed and linear phases, which corresponds to three phases of the Generalized REM, well explained in (Xxxxxxx and Xxxxxxx, 1986, Section 3). Its behavior is visualized in Fig. 5. We make the following brief observations, explaining the combinatorial meaning for an approximate learning process: ^ — The normalized eLPA in (33) depends on the temperature by the product βσ, which pronounces the fact that the reference scale for the temperature of Xxxxx posteriors is adjusted by the amount of signal in data. — The noise-to-signal ratio γ p√lays the crucial role. For a fixed signal σ, the optimal temperature β^∗ grows to 2/σ, as the noise-to-signal vanishes (γ → 0, i.e. σ˜ σ). This behavior supports our intuition that a posterior adapted to ^ the signal variance is better to choose in the absence of noise. Optimal β∗ is located in the so called retrieval phase. Figure 5: Behavior of the expected log-posterior agreement (4). — The high temperature phase β^ → 0 results in low eLPA meaning that ^ → ∞ informative solutions cannot be found by sampling from Xxxxx posteriors (which are too “broad”) in this phase.
Discussion of the results. In Figure 40a all the tested models during the inversion are depicted (gray lines). The best fit model (white line) and the models lying inside the 10% range of the minimum cost (black lines) function are highlighted. The agreement between experimental and theoretical Rayleigh wave dispersion curves (Figure 40c-d) is good and, considering the wavelengths related to the dispersion curve frequency range, the VS profile between 5 to about 100 metres is likely to be well constrained. Therefore, since below this depth the profile is constrained by the H/V curve alone, we prefer to show, in Figure 40a and Table 6, the VS profile only within the depth range were both curves contribute to the inversion.
a) b) c)
Discussion of the results. Table 3.1 summarizes the outcomes of all the experimental techniques adopted to characterize traps in our devices. When DLTS is applied to diodes (that do not have the GaN cap layer), only traps B and C are detected, while trap A is not. This suggests that trap A is located in the cap layer or at the device surface, while traps B and C are located in the device bulk. On the other hand, the remaining three techniques, that is current mode DLTS on FAT- HEMTs, low-frequency transconductance dispersion measurements and temperature- dependent gate-lag measurements (the latter two methods being applied to short-gate HEMTs), provide the same indications for the three traps. Now, for trap A all techniques consistently indicate that it must be located in the cap or at the device surface. For traps B and C, on the contrary, we got inconsistent indications, with the comparison between DLTS on FAT-HEMTs and diodes pointing at bulk localization, while current mode DLTS, transconductance and gate lag methods suggesting surface localization. DLTS on diodes NOT detected Detected Detected DLTS (FAT-HEMTs) Detected Detected Detected I-DLTS (FAT-HEMTs) “Hole-like” peak “Hole-like” peak “Hole-like” peak gm(f) N.A. Negative dispersion N.A. Gate lag measurements Increasing ID transient Increasing ID transient Increasing ID transient Location Surface, cap Inconsistent indications Inconsistent indications In order to investigate these contrasting evidences, the drain-current (ID) transients were measured at room temperature, where trap B dominates (see Figure 3.2), by using different VGS test pulses. As a matter of fact, as long as VGS is pulsed from 0 V to a negative value (VGS,off) higher than the pinch-off voltage, the ID transient is of the type that is generally attributed to surface traps (see Figure 3.7). However, when VGS,off approaches the pinch-off voltage (that is about 4.8 V in the devices under test), the shape of the ID transient changes to that characteristic of a bulk trap (Figure 3.8). In this regard, some questions arise: can the same trap be at the origin of the two opposite ID transient behaviors? and why? and where must this trap be located? To try to answer to these questions, two-dimensional numerical simulations were carried out by using a commercial code. VGS VDS = 1.5V 0.35 0.34 a.u. 0.33 0.32 0.31 0.3 0 0.01 0.02 0.03 0.04 0.05 Time (s) Figure 3.7: Experimental drain-current transients in the FAT-HEMTs for a VGS pulse from 0V to -3V at a VDS of 1.5V. Ex...
