README.md 1.21 KB
Newer Older
Daniel García's avatar
Daniel García committed
1
2
3
All code relating to the specific experiments conducted on the data published in https://lbd.udc.es/research/real-life-HAR-dataset.

Note that each script is coded in a Slurm way to be executed as a job array, limited to 250 tasks at a time (in our case). The parameters for every script would be as follows: 
Daniel García's avatar
Daniel García committed
4
5
- 'windowSize_overlap_dataset' featureSet slurmArrayTaskId [firstOrLast250]

Daniel García's avatar
Daniel García committed
6
7
8
Each parameter could take the following values:
- WindowSize = [20, 30, 40, 50, 60, 70, 80, 90]
- Overlap = [19.0, 29.0, 39.0, 49.0, 59.0, 69.0, 79.0, 89.0]
Daniel García's avatar
Daniel García committed
9
    - Each value corresponds to one and only one of the previous window sizes.
Daniel García's avatar
Daniel García committed
10
- Dataset = [1, 2] 
Daniel García's avatar
Daniel García committed
11
12
    - 2 - Complete dataset (all sensors)
    - 1 - Dataset without gyroscope
Daniel García's avatar
Daniel García committed
13
- FeatureSet = [0, 1] 
Daniel García's avatar
Daniel García committed
14
15
    - 0 - Initial feature set (basic)
    - 1 - Proposed feature set (more complex)
Daniel García's avatar
Daniel García committed
16
- SlurmArrayTaskId = [1-10] or [1-250]
Daniel García's avatar
Daniel García committed
17
18
    - Each task corresponds to one of the initial folds done in the experiments (1-10).
    - For the non-deterministic algorithms (RF and MLP, in this case), we need to do at least 50 experiments to bypass its random behaviour (1-250, two times).
Daniel García's avatar
Daniel García committed
19
- FirstOrLast250 = [0, 1] 
Daniel García's avatar
Daniel García committed
20
    - This is only needed for the non-deterministic algorithms, due to the 250 task limit we had.