def main(session, noun=10, verb=2):
raw = common.load_input(2, session)
lines = parse(raw)
lines[1] = noun
lines[2] = verb
return run(lines)
def main(session):
raw = common.load_input(1, session)
lines = list(map(int, raw.splitlines()))
part1 = sum(map(fuel, lines))
part2 = sum(map(total_fuel_recursive, lines))
print(f"Part 1: {part1}")
print(f"Part 2: {part2}")
def main(session=None):
raw = common.load_input(3, session)
lines = raw.splitlines()
wires = list(map(wire2multiline, [line.split(',') for line in lines]))
intersections = wires[0].intersection(wires[1])
distances = list(map(manhattan_distance, intersections))
distances.sort()
print (f"Part 1: {distances[1]}")
distances2 = list(map(wire_length_gen(wires), intersections[1:]))
distances2.sort()
print (f"Part 2: {distances2[1]}")
save_to_value_store(value_name, value)
p = publish_value
t = start_stop_timer
outcome_diagnosis_pattern = ("I") #|| input
use_death_as_outcome = False #|| input
followup_period_years = 5 #|| input
#load_input_run_name = "T2D_timesplit__Stroke__v14"
load_input_run_name = "T2D_timesplit__CKD_N17-N19__v14"
load_input_shared_storage_name = "T2D__to__OUTCOMES_MIX_v14_1__SHARED_DATA"
subpop_lbnrs = load_input(load_input_run_name, "subpop_lbnrs", load_input_shared_storage_name) #|| input
first_subpop_event_df = load_input(load_input_run_name, "first_subpop_event_df", load_input_shared_storage_name) #|| input
subpop_cpr_df = load_input(load_input_run_name, "subpop_cpr_df", load_input_shared_storage_name) #|| input
subpop_birthplace_df = load_input(load_input_run_name, "subpop_birthplace_df", load_input_shared_storage_name) #|| input
subpop_death_causes_df = load_input(load_input_run_name, "subpop_death_causes_df", load_input_shared_storage_name) #|| input
subpop_sks_df = load_input(load_input_run_name, "subpop_sks_df", load_input_shared_storage_name) #|| input
subpop_diag_df = load_input(load_input_run_name, "subpop_diag_df", load_input_shared_storage_name) #|| input
subpop_prescriptions_df = load_input(load_input_run_name, "subpop_prescriptions_df", load_input_shared_storage_name) #|| input
subpop_ssr_df = load_input(load_input_run_name, "subpop_ssr_df", load_input_shared_storage_name) #|| input
current_addresses_df = load_input(load_input_run_name, "current_addresses_df", load_input_shared_storage_name) #|| input
past_addresses_df = load_input(load_input_run_name, "past_addresses_df", load_input_shared_storage_name) #|| input
past_archive_addresses_df = load_input(load_input_run_name, "past_archive_addresses_df", load_input_shared_storage_name) #|| input
#load_input_run_name = "T2D_timesplit__Stroke__v14"
load_input_shared_storage_name = "T2D__to__OUTCOMES_MIX_v14_1__SHARED_DATA"
#%%
#||
action_name = "Describe Fit Results"
action_description = ""
action_input = {"fit_results", "trainX", "testX", "trainY", "testY", "patient_feature_vector___index__to__feature__list",
"testX_index__to__patient_feature_vector_indexes", "patient_feature_vector___index__to__lbnr__list", "subpop_basic_df"}
action_output = {"model_type_summary"}
#||
#%%
#fit_results = load_input(load_input_run_name, "ExGradientBoost_count__fit_results", load_input_shared_storage_name) #|| input
#fit_results = load_input(load_input_run_name, "LR_aux__fit_results", load_input_shared_storage_name) #|| input
fit_results = load_input(load_input_run_name, "ExGradientBoost_count__fit_results", load_input_shared_storage_name) #|| input
#trainX = load_input(load_input_run_name, "aux__trainX", load_input_shared_storage_name) #|| input
trainX = load_input(load_input_run_name, "count__trainX", load_input_shared_storage_name) #|| input
trainY = load_input(load_input_run_name, "trainY", load_input_shared_storage_name) #|| input
#testX = load_input(load_input_run_name, "aux__testX", load_input_shared_storage_name) #|| input
testX = load_input(load_input_run_name, "count__testX", load_input_shared_storage_name) #|| input
testY = load_input(load_input_run_name, "testY", load_input_shared_storage_name) #|| input
#valX = load_input(load_input_run_name, "aux__valX", load_input_shared_storage_name) #|| input
valX = load_input(load_input_run_name, "count__valX", load_input_shared_storage_name) #|| input
valY = load_input(load_input_run_name, "valY", load_input_shared_storage_name) #|| input
trainX_index__to__patient_feature_vector_indexes = load_input(load_input_run_name, "trainX_index__to__patient_feature_vector_indexes", load_input_shared_storage_name) #|| input
testX_index__to__patient_feature_vector_indexes = load_input(load_input_run_name, "testX_index__to__patient_feature_vector_indexes", load_input_shared_storage_name) #|| input
valX_index__to__patient_feature_vector_indexes = load_input(load_input_run_name, "valX_index__to__patient_feature_vector_indexes", load_input_shared_storage_name) #|| input
patient_feature_vector___index__to__lbnr__list = load_input(load_input_run_name, "patient_feature_vector___index__to__lbnr__list", load_input_shared_storage_name) #|| input
patient_feature_vector___index__to__feature__list = load_input(load_input_run_name, "patient_feature_vector___index__to__feature__list", load_input_shared_storage_name) #|| input
pd.options.display.max_rows = 30
#%%
#Additional data that i had to recalculate for publication purposes
#Print poulation sizes including followup filtering
from datetime import datetime, timedelta
last_calendar_day_of_data = datetime(2016, 1, 1)
last_day_of_data = last_calendar_day_of_data - timedelta(days=365 * 5)
for i in range(5):
load_input_run_name = f"T2D_{i}_timesplit__MI__v14" if i != 0 else "T2D__timesplit__MI__v14"
load_input_shared_storage_name = f"T2D_{i}__to__OUTCOMES_MIX_v14_1__SHARED_DATA" if i != 0 else "T2D__to__OUTCOMES_MIX_v14_1__SHARED_DATA"
subpop_cpr_df = load_input(load_input_run_name, "subpop_cpr_df",
load_input_shared_storage_name) #|| input
first_subpop_event_df = load_input(
load_input_run_name, "first_subpop_event_df",
load_input_shared_storage_name) #|| input
filtered_first_subpop_event_df = first_subpop_event_df[
first_subpop_event_df["first_subpop_event"] < last_day_of_data]
filtered_cpr_df = pd.merge(subpop_cpr_df,
filtered_first_subpop_event_df,
left_index=True,
right_index=True,
how="right")
print("{}: # {}\t%W {:.3f}\tage {:.3f}".format(
i, filtered_cpr_df.shape[0], 100 *
float(filtered_cpr_df[filtered_cpr_df["C_KON"] == "K"].shape[0]) /
#!/usr/bin/env python3
import numpy as np
from common import load_input
from common import sigmoid
inputs, labels, labels_1hot, (K, m, n) = load_input('predict.csv')
weights_file = 'weights.csv'
thetas = np.loadtxt(weights_file, delimiter=',')
print(f'loaded weights from {weights_file}')
for i in range(len(inputs)):
image = inputs[i, :]
image = image.reshape(image.shape[0], 1)
label = labels[i]
print(f'\nimage {i} has label {label}')
ps = sigmoid(thetas.dot(image))
top_i = np.argmax(ps)
top = np.squeeze(ps[top_i])
print(f'prediction: {top_i} with {top * 100:.3f}%')
for k in range(len(ps)):
p = np.squeeze(ps[k])
print(f'\tP({k}) = {p * 100:7.3f}%')
#!/usr/bin/env python3
import numpy as np
from common import load_input
from common import sigmoid
inputs, labels, labels_1hot, (K, m, n) = load_input('train.csv')
thetas = np.random.rand(K, n)
alpha = 1e-3
iters = int(20_000)
X = inputs
for k in range(K):
print(f'train classifier {k} with {m} examples for {iters} iterations')
theta = thetas[k].reshape(n, 1)
Y = labels_1hot[k].reshape(m, 1)
for i in range(iters):
z = X.dot(theta)
a = sigmoid(z)
diff = a - Y
grad = X.T.dot(diff)
theta -= (alpha / m) * grad
if i == iters - 1:
pass
# TODO: calculate cost