|
DETAILS
|
TorchSVM
|
TorchSVM
MODEL
|
| |
| |
Support Vector Machine |
| |
| |
+ |
degree
|
INT
|
1
|
'-1'
|
LOCK
|
if positive, a polynomial kernel [(a xy + b)^d] with the specified degree is used, otherwise a gaussian kernel is used
|
| |
+ |
a
|
FLOAT
|
1
|
'1.00000'
|
LOCK
|
constant a in the polynomial kernel
|
| |
+ |
b
|
FLOAT
|
1
|
'1.00000'
|
LOCK
|
constant b in the polynomial kernel
|
| |
+ |
std
|
FLOAT
|
1
|
'10.00000'
|
LOCK
|
the std parameter in the gaussian kernel [exp(-|x-y|^2/std^2)]
|
| |
+ |
iter
|
FLOAT
|
1
|
'0.00000'
|
LOCK
|
minimal number of iterations before shrinking
|
| |
+ |
acc
|
FLOAT
|
1
|
'0.01000'
|
LOCK
|
end accuracy
|
| |
+ |
c
|
FLOAT
|
1
|
'100.00000'
|
LOCK
|
trade off cst between error/margin
|
| |
+ |
cache
|
FLOAT
|
1
|
'50.00000'
|
LOCK
|
cache size in Mo
|
| |
|
TorchKNN
|
TorchKNN
MODEL
|
| |
| |
K-Nearest Neighbor |
| |
| |
+ |
k
|
UINT
|
1
|
'3'
|
LOCK
|
k neighbours
|
| |
|
TorchKMeans
|
TorchKMeans
MODEL
|
| |
| |
K-Means |
| |
| |
+ |
clusteres
|
INT
|
1
|
'3'
|
LOCK
|
number of clusters
|
| |
+ |
prior
|
FLOAT
|
1
|
'0.00100'
|
LOCK
|
prior weights (only initialization)
|
| |
+ |
accuracy
|
FLOAT
|
1
|
'0.00001'
|
LOCK
|
accuracy
|
| |
+ |
iter_kmeans
|
INT
|
1
|
'25'
|
LOCK
|
max iterations during k-mean (only initalization)
|
| |
+ |
threshold
|
FLOAT
|
1
|
'0.00100'
|
LOCK
|
accuracy
|
| |
|
TorchHMM
|
TorchHMM
MODEL
|
| |
| |
Hidden Markov Model |
| |
| |
+ |
single
|
BOOL
|
1
|
'false'
|
LOCK
|
use single ergodic hmm with #states = #classes (overrides -states and -connectivity option)
|
| |
+ |
connectivity
|
INT
|
1
|
'0'
|
LOCK
|
connectivity (linear=0, left-right=1, ergodic=2)
|
| |
+ |
states
|
INT
|
1
|
'3'
|
LOCK
|
number of states
|
| |
+ |
gaussians
|
INT
|
1
|
'3'
|
LOCK
|
number of gaussians
|
| |
+ |
prior
|
FLOAT
|
1
|
'0.00100'
|
LOCK
|
prior weights (only initialization)
|
| |
+ |
accuracy
|
FLOAT
|
1
|
'0.00001'
|
LOCK
|
accuracy
|
| |
+ |
iter_kmeans
|
INT
|
1
|
'25'
|
LOCK
|
max iterations during k-mean (only initalization)
|
| |
+ |
threshold
|
FLOAT
|
1
|
'0.00100'
|
LOCK
|
accuracy
|
| |
+ |
iter_gmm
|
INT
|
1
|
'25'
|
LOCK
|
max iterations during training (only retraining)
|
| |
+ |
retrain
|
INT
|
1
|
'2'
|
LOCK
|
number of retrains
|
| |
|
TorchGMM
|
TorchGMM
MODEL
|
| |
| |
Gaussian Mixture Model |
| |
| |
+ |
gaussians
|
INT
|
1
|
'3'
|
LOCK
|
number of gaussians
|
| |
+ |
prior
|
FLOAT
|
1
|
'0.00100'
|
LOCK
|
prior weights (only initialization)
|
| |
+ |
accuracy
|
FLOAT
|
1
|
'0.00001'
|
LOCK
|
accuracy
|
| |
+ |
iter_kmeans
|
INT
|
1
|
'25'
|
LOCK
|
max iterations during k-mean (only initalization)
|
| |
+ |
threshold
|
FLOAT
|
1
|
'0.00100'
|
LOCK
|
accuracy
|
| |
+ |
iter_gmm
|
INT
|
1
|
'25'
|
LOCK
|
max iterations during training (only retraining)
|
| |
+ |
retrain
|
INT
|
1
|
'2'
|
LOCK
|
number of retrains
|
| |
|