2.1.0
User Documentation for Apache MADlib
crf.sql_in File Reference

SQL functions for conditional random field. More...

Functions

float8 [] lincrf_lbfgs_step_transition (float8[], float8[], float8[], float8[], float8, float8, float8[])
 
float8 [] lincrf_lbfgs_step_merge_states (float8[] state1, float8[] state2)
 
float8 [] lincrf_lbfgs_step_final (float8[] state)
 
float8 internal_lincrf_lbfgs_converge (float8[] state)
 
lincrf_result internal_lincrf_lbfgs_result (float8[] state)
 
aggregate float8 [] lincrf_lbfgs_step (float8[], float8[], float8[], float8, float8, float8[])
 
aggregate anyarray array_union (anyarray)
 
integer compute_lincrf (text source, text sparse_R, text dense_M, text sparse_M, text featureSize, integer tagSize, integer maxNumIterations)
 
text lincrf_train (text train_feature_tbl, text train_featureset_tbl, text label_tbl, text crf_stats_tbl, text crf_weights_tbl, integer max_iterations)
 Compute linear-chain crf coefficients and diagnostic statistics. More...
 
text lincrf_train (text train_feature_tbl, text train_featureset_tbl, text label_tbl, text crf_stats_tbl, text crf_weights_tbl)
 

Detailed Description

Date
July 2012
See also
For a brief introduction to conditional random field, see the module description Conditional Random Field.

Function Documentation

◆ array_union()

aggregate anyarray array_union ( anyarray  )

◆ compute_lincrf()

integer compute_lincrf ( text  source,
text  sparse_R,
text  dense_M,
text  sparse_M,
text  featureSize,
integer  tagSize,
integer  maxNumIterations 
)

◆ internal_lincrf_lbfgs_converge()

float8 internal_lincrf_lbfgs_converge ( float8 []  state)

◆ internal_lincrf_lbfgs_result()

lincrf_result internal_lincrf_lbfgs_result ( float8 []  state)

◆ lincrf_lbfgs_step()

aggregate float8 [] lincrf_lbfgs_step ( float8  [],
float8  [],
float8  [],
float8  ,
float8  ,
float8  [] 
)

◆ lincrf_lbfgs_step_final()

float8 [] lincrf_lbfgs_step_final ( float8 []  state)

◆ lincrf_lbfgs_step_merge_states()

float8 [] lincrf_lbfgs_step_merge_states ( float8 []  state1,
float8 []  state2 
)

◆ lincrf_lbfgs_step_transition()

float8 [] lincrf_lbfgs_step_transition ( float8  [],
float8  [],
float8  [],
float8  [],
float8  ,
float8  ,
float8  [] 
)

◆ lincrf_train() [1/2]

text lincrf_train ( text  train_feature_tbl,
text  train_featureset_tbl,
text  label_tbl,
text  crf_stats_tbl,
text  crf_weights_tbl,
integer  max_iterations 
)
Parameters
sourceName of the source relation containing the training data
sparse_RName of the sparse single state feature column (of type DOUBLE PRECISION[])
dense_MName of the dense two state feature column (of type DOUBLE PRECISION[])
sparse_MName of the sparse two state feature column (of type DOUBLE PRECISION[])
featureSizeName of feature size column (of type DOUBLE PRECISION)
tagSizeThe number of tags in the tag set
featuresetThe unique feature set
crf_featureThe Name of output feature table
maxNumIterationsThe maximum number of iterations
Returns
a composite value:
  • coef FLOAT8[] - Array of coefficients, \( \boldsymbol c \)
  • log_likelihood FLOAT8 - Log-likelihood \( l(\boldsymbol c) \)
  • num_iterations INTEGER - The number of iterations before the algorithm terminated

    A 'crf_feature' table is used to store all the features and corresponding weights
Note
This function starts an iterative algorithm. It is not an aggregate function. Source and column names have to be passed as strings (due to limitations of the SQL syntax).

◆ lincrf_train() [2/2]

text lincrf_train ( text  train_feature_tbl,
text  train_featureset_tbl,
text  label_tbl,
text  crf_stats_tbl,
text  crf_weights_tbl 
)