forked from mirrors/nixpkgs
41 lines
1.3 KiB
Python
41 lines
1.3 KiB
Python
from pyspark.sql import Row, SparkSession
|
|
from pyspark.sql import functions as F
|
|
from pyspark.sql.functions import udf
|
|
from pyspark.sql.types import *
|
|
from pyspark.sql.functions import explode
|
|
|
|
def explode_col(weight):
|
|
return int(weight//10) * [10.0] + ([] if weight%10==0 else [weight%10])
|
|
|
|
spark = SparkSession.builder.getOrCreate()
|
|
|
|
dataSchema = [
|
|
StructField("feature_1", FloatType()),
|
|
StructField("feature_2", FloatType()),
|
|
StructField("bias_weight", FloatType())
|
|
]
|
|
|
|
data = [
|
|
Row(0.1, 0.2, 10.32),
|
|
Row(0.32, 1.43, 12.8),
|
|
Row(1.28, 1.12, 0.23)
|
|
]
|
|
|
|
df = spark.createDataFrame(spark.sparkContext.parallelize(data), StructType(dataSchema))
|
|
|
|
normalizing_constant = 100
|
|
sum_bias_weight = df.select(F.sum('bias_weight')).collect()[0][0]
|
|
normalizing_factor = normalizing_constant / sum_bias_weight
|
|
df = df.withColumn('normalized_bias_weight', df.bias_weight * normalizing_factor)
|
|
df = df.drop('bias_weight')
|
|
df = df.withColumnRenamed('normalized_bias_weight', 'bias_weight')
|
|
|
|
my_udf = udf(lambda x: explode_col(x), ArrayType(FloatType()))
|
|
df1 = df.withColumn('explode_val', my_udf(df.bias_weight))
|
|
df1 = df1.withColumn("explode_val_1", explode(df1.explode_val)).drop("explode_val")
|
|
df1 = df1.drop('bias_weight').withColumnRenamed('explode_val_1', 'bias_weight')
|
|
|
|
df1.show()
|
|
|
|
assert(df1.count() == 12)
|