References in periodicals archive ?
For each subject, the nEMG data were separated into patterns of synergies and muscle weightings using an NNMF algorithm .
The NNMF algorithm was initialized with two random matrices of activation patterns and weightings.
The NNMF analyses were performed with the output restricted to one, two, three, four, or five synergies, with no a priori assumptions about the adequate number of synergies.
First, interclass correlation coefficients ([ICC.sub.(1,1)]) were calculated between the two trials of the first measurements to investigate the test-retest reliability of the number of synergies indicated by NNMF. Second, the paired t-test and Wilcoxon signed-rank test were used to examine the differences in clinical parameters, gait kinematics, and the number of synergies between the two measurements.
Our present study clarified the longitudinal change in muscle synergy calculated using NNMF for patients after subacute stroke.
Algorithm MAP (mean averge precision) standard deviation (AP) NNMF  0.583 0.015 LMNN  0.586 0.020 LTML  0.597 0.011 FESS  0.592 0.017 FKL  0.601 0.014 LDA-FEK (ours) 0.613 0.013 Table 3.
From Table 2, we find that Algorithm 3 is still better than NNMF for most of the test problems.
It is obvious that average numbers of iterations and the minimum value of objective function value for Algorithm 3 are smaller than those for NNMF, CPU time of Algorithm 3 is less than that of NNMF, and the speed of convergence of Algorithm 3 and NNMF is very close.
In Figure 1, we compare the performance of the proposed method with NNMF, respectively.
And the numbers of iterations for Algorithm 3 are smaller than that for NNMF.
It can be clearly found that, FK, FESS and PD outperform NNMF and LMNN.
Algorithm MAP (mean averge precision) NNMF  0.583 LMNN  0.586 FK  0.588 FESS  0.590 PD  0.593 DFK  0.606 PLSA-FK (ours) 0.619 Table 4.
Acronyms browser ?
Full browser ?
- NNN medium