FLOOXS » TclLib » Analysis
17 documented proc(s) in TclLib/Analysis/.
__grad_slurm · __run_errorfxn · __updateLearningRates · __updateParams · __wait_srun · __WT · ApplyMask · grad · old_grad · optimize · optimize · optimize · penalty_range · run_errorfxn_yi · updateLearningRates · updateParams · WT
errorfxn: name of the error function that will be run errorPath: the path to the file containing the error function. It has be be sources params: the input parameters for the error function grad_type: the type of gradient that will be calculated (forwards, backwards, or central)
errorfxn: name of the error function that will be run errorPath: the path to the file containing the error function. It has be be sources params: the input parameters for the error function o_file: the path to the output file where the results from the error function will be stored print_str: the string that will be combined be combined with the error fucntion results and printed to the output file
updates the learning rate based on how much the previous parameters changed divided by the gradient difference between the two points
for each parameter, if the partial derivative of the error function with respect to that parameter is positive, decrease the value of the parameter, if it is negative, increase the value of the parameter the magnitude of the parameter change is dependent on the learning rate and if it is too grate, it is reduced to the damp value
job_ID: the ID number of the current job
### LARS test procs #### expr ([lindex y_del $i] - $y0)/($delta) These are to optimization procs
# Detect parameter mask
errorfxn: name of the error function that will be run errorPath: the path to the file containing the error function. It has be be sources params: the input parameters for the error function job_ID: the ID number of the current job grad_type: the type of gradient that will be calculated (forwards, backwards, or central)
## actual optimize procs ####
declares the error function, the initial input parameters, the number of iteration to be used, and the damping level
declares the error function, the initial input parameters, the number of iteration to be used, and the damping level
{ errorFxn params {options {}} }
declares the error function, the initial input parameters, the number of iteration to be used, and the damping level
{ errorFxn params {options {}} }
SLURM dispatch via __grad_slurm is preserved; that proc lives in legacy/
optimize_slurm.tcl and must be sourced manually when running under SLURM.
square penalty
errorfxn: name of the error function that will be run errorPath: the path to the file containing the error function. It has be be sources params: the input parameters for the error function o_file: the path to the output file where the results from the error function will be stored print_str: the string that will be combined be combined with the error fucntion results and printed to the output file
updates the learning rate based on how much the previous parameters changed divided by the gradient difference between the two points
for each parameter, if the partial derivative of the error function with respect to that parameter is positive, decrease the value of the parameter, if it is negative, increase the value of the parameter the magnitude of the parameter change is dependent on the learning rate and if it is too grate, it is reduced to the damp value
### LARS test procs #### expr ([lindex y_del $i] - $y0)/($delta) These are to optimization procs