| | | |
Best hyper parameters (with respect to SNR on leave-1 training dataset)
|
Performed on training set:
|
Performed on test set:
| |
|---|
|
Dynamics
|
Normalization
|
Optimization
|
Gamma (state-space coefficient)
|
Tau (kinetic time constant)
|
Lambda (regularization parameter)
|
SNR (in dB) on leave-1 training dataset
|
percentage of correct signs on leave-1 test dataset
|
Reference
|
|---|
|
Kinetic
|
MAS5
|
Gradient
|
1
|
3
|
0.0001
|
32.4
|
68%
|
This work
|
|
Kinetic
|
MAS5
|
LARS
|
0.1
|
3
|
0.1
|
32.4
|
74%
|
This work
|
|
Kinetic
|
MAS5
|
LARS
|
0
|
3
|
0.05
|
32.1
|
74%
|
[33]
|
|
kinetic
|
MAS5
|
Elastic Nets
|
0
|
3
|
0.05
|
32.1
|
74%
|
[35]
|
|
Brownian
|
MAS5
|
Gradient
|
0
|
NA
|
0.005
|
32.1
|
66%
|
[34]
|
|
Brownian
|
MAS5
|
LARS
|
0
|
NA
|
0.05
|
32.1
|
63%
|
[34]
|
|
Brownian
|
MAS5
|
Elastic Nets
|
0
|
NA
|
0.05
|
32.1
|
63%
|
[34]
|
|
Naïve trend prediction
|
MAS5
| |
NA
|
NA
|
NA
|
NA
|
52%
| |
- We compared our SSM-based technique (with a non-zero SSM parameter gamma) to previously published algorithms for learning gene regulation networks by enforcing gamma = 0 (see Materials and methods). We notice that the LARS algorithm [42], used in the Inferelator by Bonneau et al. [32, 33], as well as Elastic Nets [35, 43], obtain a slightly worse quality of fit (signal-to-noise ratio (SNR), in dB) than when combined with our state-space modeling for the same leave-out-last (leave-1) performance as our SSM plus LARS. Not using an mRNA degradation term, as in Wang et al. [34], degrades the leave-out-last performance.
