def compute_conditional_probability_distribution(variables, data, motor, conditioned):
"""
"""
freq, n = DistributionComputer.compute_frequency_table(variables, data, motor)
# Total number of occurences with only exploration variables
totals = SparseTable({k: v for k, v in variables.iteritems()
if k not in conditioned})
jpd = SparseConditionalTable(variables, {k: v for k, v in variables.iteritems() if k not in conditioned})
# Make conditional on exploration variables
# i.e. divide by total number of
for entry in freq.get_nonzero_entries():
filtered = {k: v for k, v in entry.iteritems() if k not in conditioned}
# Add the values to increase the total
totals.set_value(filtered, totals.get_value(filtered) + freq.get_value(entry))
# Make conditional table by dividing by subtotals
for entry in freq.get_nonzero_entries():
filtered = {k: v for k, v in entry.iteritems() if k not in conditioned}
# P(M|R) = F(M&R) / F(R)
f_r = float(totals.get_value(filtered))
jpd.set_value(entry, 0 if f_r == 0 else freq.get_value(entry) / f_r)
return jpd
def compute_frequency_table(variables, data, motor):
"""
Parameters
----------
variables : [string]
data : Data
motor : [string]
"""
freq = SparseTable(variables)
first = 2
last = data.last_time() + 1
n = 0
for t_i in range(first, last):
n += 1
freq.inc_value(data.get_entries_previous_current(t_i, variables.keys(), motor))
return freq, n
def compute_frequency_table(variables, data, motor):
"""
Parameters
----------
variables : [string]
data : Data
motor : [string]
"""
freq = SparseTable(variables)
first = 2
last = data.last_time() + 1
n = 0
for t_i in range(first, last):
n += 1
freq.inc_value(
data.get_entries_previous_current(t_i, variables.keys(),
motor))
return freq, n
def compute_conditional_probability_distribution(variables, data, motor,
conditioned):
"""
"""
freq, n = DistributionComputer.compute_frequency_table(
variables, data, motor)
# Total number of occurences with only exploration variables
totals = SparseTable(
{k: v
for k, v in variables.iteritems() if k not in conditioned})
jpd = SparseConditionalTable(
variables,
{k: v
for k, v in variables.iteritems() if k not in conditioned})
# Make conditional on exploration variables
# i.e. divide by total number of
for entry in freq.get_nonzero_entries():
filtered = {
k: v
for k, v in entry.iteritems() if k not in conditioned
}
# Add the values to increase the total
totals.set_value(
filtered,
totals.get_value(filtered) + freq.get_value(entry))
# Make conditional table by dividing by subtotals
for entry in freq.get_nonzero_entries():
filtered = {
k: v
for k, v in entry.iteritems() if k not in conditioned
}
# P(M|R) = F(M&R) / F(R)
f_r = float(totals.get_value(filtered))
jpd.set_value(entry,
0 if f_r == 0 else freq.get_value(entry) / f_r)
return jpd
