def create_submission(model, datasets, raw_datasets):
submission_parts = []
for ds, raw_ds in zip(datasets, raw_datasets):
ds_predictions = get_predictions(ds, model)
submission_part = predictions_to_submission(ds_predictions, raw_ds)
submission_parts.append(submission_part)
submission = pd.concat(submission_parts, ignore_index=True)
return submission
def test(mhc, data, model, model_path):
'''
Training protocol
'''
# print out options
print('Testing\nMHC: %s\nData: %s\nModel: %s\nSave path: %s' %
(mhc, data, model, model_path))
# load training
test_data = Dataset.from_csv(filename=data,
sep=',',
allele_column_name='mhc',
peptide_column_name='peptide',
affinity_column_name='IC50(nM)')
# apply cut/pad or mask to same length
if 'lstm' in model or 'gru' in model:
test_data.mask_peptides()
else:
test_data.cut_pad_peptides()
# get the allele specific data
mhc_test = test_data.get_allele(mhc)
# define model
if model == 'fc':
model = models.mhcnuggets_fc()
elif model == 'gru':
model = models.mhcnuggets_gru()
elif model == 'lstm':
model = models.mhcnuggets_lstm()
elif model == 'chunky_cnn':
model = models.mhcnuggets_chunky_cnn()
elif model == 'spanny_cnn':
model = models.mhcnuggets_spanny_cnn()
# compile model
model.load_weights(model_path)
model.compile(loss='mse', optimizer=Adam(lr=0.001))
# get tensorized values for training
test_peptides, test_continuous, test_binary = mhc_test.tensorize_keras(
embed_type='softhot')
# test
preds_continuous, preds_binary = get_predictions(test_peptides, model)
test_auc = roc_auc_score(test_binary, preds_continuous)
test_f1 = f1_score(test_binary, preds_binary)
test_ktau = kendalltau(test_continuous, preds_continuous)[0]
print('Test AUC: %.4f, F1: %.4f, KTAU: %.4f' %
(test_auc, test_f1, test_ktau))
def main():
if len(sys.argv) <= 2:
print("Please provide the inputfile(realtimedata)", len(sys.argv))
exit(-1)
window = 20
window2 = 10
user_rating = 5
chances = 0.2
inputfile = sys.argv[1]
out_dir = sys.argv[1].split('.')[0] + "_filtered"
if len(sys.argv) == 3:
print("User past rating out of 10", int(sys.argv[2]))
user_rating = int(sys.argv[2])
baseline_dataset = load_my_data("basedata/main.csv", ",")
baseline_dataset = preprocess_data(baseline_dataset, window, window2)
#print(baseline_dataset)
m_mdls = mdl.train_model(baseline_dataset)
print("DONE TRAINING MODELS WITH BASELINE DATASET")
#os.mkdir(out_dir)
filename = inputfile
dataset1 = load_my_data(inputfile, ',')
sliding_samples = 120
rows = dataset1.shape[0]
for timestep in range(int(rows / sliding_samples) * sliding_samples):
start = timestep
end = timestep + sliding_samples
print("PROCESSING WINDOW FROM ", start, " UNTIL ", end)
if end > rows:
end = rows - 1
if end - start + 1 < 12:
timestep += sliding_samples
continue
dset1 = dataset1[start:end]
dset1 = preprocess_data(dset1, window, window2)
#print(dset1)
out_lbls, prob = mdl.get_predictions(dset1, m_mdls)
chances, user_rating = get_accident_probability(
user_rating, out_lbls, prob, chances)
#write_to_file(out_dir, inputfile.split('.')[0], dataset1)
timestep += sliding_samples
print("CHANCES OF ACCIDENT:", chances, " NEW USER RATING:",
round(user_rating))
def calculate_relation(mhc, data, model, weights_dir):
'''
Training protocol
'''
print('Calculating tuning MHC for %s' % mhc)
relations_dict = {}
# get the allele specific data
mhc_data = data.get_allele(mhc)
train_peptides, train_continuous, train_binary = mhc_data.tensorize_keras(embed_type='softhot')
best_mhc = ''
best_auc = 0
num_mhc = len(mhc_data.peptides)
for tuning_mhc in sorted(set(data.alleles)):
# don't want to tune with ourselves
if mhc == tuning_mhc:
continue
# define the path to save weights
try:
model_path = os.path.join(weights_dir, tuning_mhc + '.h5')
model.load_weights(model_path)
except IOError:
continue
preds_continuous, preds_binary = get_predictions(train_peptides, model)
try:
auc = roc_auc_score(train_binary, preds_continuous)
#print ('MHC: %s, AUC: %.4f, F1: %.4f, KTAU: %.4f' % (tuning_mhc,
# auc,
# f1,
# ktau))
if auc > best_auc:
best_mhc = tuning_mhc
best_auc = auc
num_tuning_mhc = len(data.get_allele(tuning_mhc).peptides)
except ValueError:
continue
return best_mhc, best_auc, num_mhc, num_tuning_mhc
def predict(model, weights_path, peptides_path):
'''
Training protocol
'''
# read peptides
peptides = [p.strip() for p in open(peptides_path)]
print('Predicting for %d peptides' % (len(peptides)))
# apply cut/pad or mask to same length
if 'lstm' in model or 'gru' in model:
normed_peptides = dataset.mask_peptides(peptides)
else:
normed_peptides = dataset.cut_pad_peptides(peptides)
# get tensorized values for prediction
peptides_tensor = dataset.tensorize_keras(normed_peptides,
embed_type='softhot')
# make model
print('Building model')
# define model
if model == 'fc':
model = models.mhcnuggets_fc()
elif model == 'gru':
model = models.mhcnuggets_gru()
elif model == 'lstm':
model = models.mhcnuggets_lstm()
elif model == 'chunky_cnn':
model = models.mhcnuggets_chunky_cnn()
elif model == 'spanny_cnn':
model = models.mhcnuggets_spanny_cnn()
model.load_weights(weights_path)
model.compile(loss='mse', optimizer=Adam(lr=0.001))
# test model
preds_continuous, preds_binary = get_predictions(peptides_tensor, model)
ic50s = [dataset.map_proba_to_ic50(p[0]) for p in preds_continuous]
for i, peptide in enumerate(peptides):
print(peptide, ic50s[i])
def run_app():
css.hide_menu()
css.limit_plot_size()
# Get cached country data
countries = fetch_country_data()
if countries.stale:
st.caching.clear_cache()
countries = fetch_country_data()
st.markdown(
body=generate_html(text=f"Corona Calculator", bold=True, tag="h1"),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
tag="h2",
text="A tool to help you visualize the impact of social distancing <br>",
),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
text="<strong>Disclaimer:</strong> <em>The creators of this application are not healthcare professionals. "
"The illustrations provided were estimated using best available data but might not accurately reflect reality.</em>",
color="gray",
font_size="12px",
),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
tag="h4",
text=f"<u><a href=\"{NOTION_MODELLING_DOC}\" target=\"_blank\" style=color:{COLOR_MAP['pink']};>"
"Methodology</a></u> <span>  </span>"
f"<u><a href=\"{MEDIUM_BLOGPOST}\" target=\"_blank\" style=color:{COLOR_MAP['pink']};>"
"Blogpost</a> </u>"
"<hr>",
),
unsafe_allow_html=True,
)
sidebar = Sidebar(countries)
country = sidebar.country
country_data = countries.country_data[country]
_historical_df = countries.historical_country_data
historical_data = _historical_df.loc[_historical_df.index == country]
number_cases_confirmed = country_data["Confirmed"]
population = country_data["Population"]
num_hospital_beds = country_data["Num Hospital Beds"]
st.subheader(f"How has the disease spread in {country}?")
st.write(
"The number of reported cases radically underestimates the true cases, because people do not show symptoms for "
"several days, not everybody gets tested, and the tests take a few days to return results. "
"The extent depends upon your country's testing strategy."
" This estimate (14% reporting) is from China ([source](https://science.sciencemag.org/content/early/2020/03/13/science.abb3221))."
)
# Estimate true cases
true_cases_estimator = models.TrueInfectedCasesModel(
constants.ReportingRate.default
)
estimated_true_cases = true_cases_estimator.predict(number_cases_confirmed)
reported_vs_true_cases(int(number_cases_confirmed), estimated_true_cases)
st.markdown(
f"Given the prevalence of the infection in your country, the probability of being infected at this time is "
f"**{estimated_true_cases / population:.3%}**. Even if you show no symptoms, the probability of being infected is "
f"**{models.get_probability_of_infection_give_asymptomatic(population, estimated_true_cases, constants.AsymptomaticRate.default):.3%}**. "
f"Note that these probabilities are on a country-wide basis and so may not apply to your situation."
)
# Plot historical data
fig = graphing.plot_historical_data(historical_data)
st.write(fig)
asymptomatic_cases_estimator = models.AsymptomaticCasesModel(
constants.AsymptomaticRate.default
)
contact_rate = sidebar.contact_rate
asymptomatic_sir_model = models.AsymptomaticSIRModel(
transmission_rate_per_contact=constants.TransmissionRatePerContact.default_per_symptom_state,
contact_rate=contact_rate,
asymptomatic_cases_model=asymptomatic_cases_estimator,
recovery_rate=constants.RecoveryRate.default,
normal_death_rate=constants.MortalityRate.default,
critical_death_rate=constants.CriticalDeathRate.default,
hospitalization_rate=constants.HospitalizationRate.default,
hospital_capacity=num_hospital_beds
)
df = models.get_predictions(
cases_estimator=true_cases_estimator,
sir_model=asymptomatic_sir_model,
num_diagnosed=number_cases_confirmed,
num_recovered=country_data["Recovered"],
num_deaths=country_data["Deaths"],
area_population=population,
)
st.subheader("How will my actions affect the spread?")
st.write(
"The critical factor for controlling spread is how many others infected people interact with each day. "
"This has a dramatic effect upon the dynamics of the disease. "
)
st.write(
"**Play with the slider to the left to see how this changes the dynamics of disease spread**"
)
df_base = df[~df.Status.isin(["Need Hospitalization"])]
base_graph = graphing.infection_graph(df_base, df_base.Forecast.max(), sidebar.contact_rate)
st.warning(graph_warning)
st.write(base_graph)
st.subheader("How will this affect my healthcare system?")
st.write(
"The important variable for hospitals is the peak number of people who require hospitalization"
" and ventilation at any one time."
)
# Do some rounding to avoid beds sounding too precise!
approx_num_beds = round(num_hospital_beds / 100) * 100
st.write(
f"Your country has around **{approx_num_beds:,}** beds. Bear in mind that most of these "
"are probably already in use for people sick for other reasons."
)
st.write(
"It's hard to know exactly how many ventilators are present per country, but there will certainly be a worldwide "
"shortage. Many countries are scrambling to buy them [(source)](https://www.reuters.com/article/us-health-coronavirus-draegerwerk-ventil/germany-italy-rush-to-buy-life-saving-ventilators-as-manufacturers-warn-of-shortages-idUSKBN210362)."
)
peak_occupancy = df.loc[df.Status == "Need Hospitalization"]["Forecast"].max()
percent_beds_at_peak = min(100 * num_hospital_beds / peak_occupancy, 100)
num_beds_comparison_chart = graphing.num_beds_occupancy_comparison_chart(
num_beds_available=approx_num_beds, max_num_beds_needed=peak_occupancy, contact_rate=sidebar.contact_rate
)
st.write(num_beds_comparison_chart)
st.markdown(
f"At peak, **{int(peak_occupancy):,}** people will need hospital beds. ** {percent_beds_at_peak:.1f}% ** of people "
f"who need a bed in hospital will have access to one given your country's historical resources. This does "
f"not take into account any special measures that may have been taken in the last few months."
)
st.subheader("How severe will the impact be?")
num_dead = df[df.Status == "Dead"].Forecast.iloc[-1]
num_recovered = df[df.Status == "Recovered"].Forecast.iloc[-1]
st.markdown(
f"If the average person in your country adopts the selected behavior, we estimate that **{int(num_dead):,}** "
f"people will die."
)
st.markdown(
f"The graph above below a breakdown of casualties and hospitalizations by age group."
)
outcomes_by_age_group = models.get_status_by_age_group(num_dead, num_recovered)
fig = graphing.age_segregated_mortality(
outcomes_by_age_group.loc[:, ["Dead", "Need Hospitalization"]], contact_rate=sidebar.contact_rate
)
st.write(fig)
st.write(
f"Parameters by age group, including demographic distribution, are [worldwide numbers](https://population.un.org/wpp/DataQuery/) "
f"so they may be slightly different in your country."
)
st.write(
f"We've used mortality rates from this [recent paper from Imperial College](https://www.imperial.ac.uk/media/imperial-college/medicine/sph/ide/gida-fellowships/Imperial-College-COVID19-NPI-modelling-16-03-2020.pdf?fbclid=IwAR3TzdPTcLiOZN5r2dMd6_08l8kG0Mmr0mgP3TdzimpqB8H96T47ECBUfTM). "
f"However, we've adjusted them according to the [maximum mortality rate recorded in Wuhan](https://wwwnc.cdc.gov/eid/article/26/6/20-0233_article)"
f" when your country's hospitals are overwhelmed: if more people who need them lack hospital beds, more people will die."
)
st.write("<hr>", unsafe_allow_html=True)
st.write(
"Like this? [Click here to share it on Twitter](https://ctt.ac/u5U39), and "
"[let us know your feedback via Google Form](https://forms.gle/J6ZFFgh4rVQm4y8G7)"
)
utils.insert_github_logo()
def run_app():
############################### Data Load ##################################
css.hide_menu()
css.limit_plot_size()
# Get cached country data
countries = _fetch_country_data()
global_data = _fetch_global_data()
if countries.stale:
st.caching.clear_cache()
countries = _fetch_country_data()
if global_data.stale:
st.caching.clear_cache()
global_data = _fetch_global_data()
################## Heading Section ####################################
utils.img_html(
alt_text='Fractal',
href='https://fractal.ai',
src=
'https://i2.wp.com/fractal.ai/wp-content/uploads/2018/02/header-black-logo.png?fit=126%2C43&ssl=1',
attributes=dict(width=125, height=43, target='_blank'))
st.markdown(
body=generate_html(text=f"Australia COVID-19 Simulator",
bold=True,
tag="h1"),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
tag="h2",
text=
"A tool to help you visualize the impact of social distancing <br>",
),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
text=
"<strong>Disclaimer:</strong> <em>The creators of this application are not healthcare professionals. "
"The illustrations provided were estimated using best available data but might not accurately reflect reality.</em>",
color="gray",
font_size="12px",
),
unsafe_allow_html=True,
)
### Add side bar
sidebar = Sidebar(countries)
###################### historical and forecast chart ##############################
### Get selected country
country = sidebar.country
country_data = countries.country_data[country]
_historical_df = countries.historical_country_data
if country == "Australia":
historical_data_custom = data_utils.make_historical_data(
_historical_df)
try:
forecasted_data = forecast_utils.get_forecasts(
country, constants.FORECAST_HORIZON)
historical_plot_df = data_utils.prep_plotting_data(
forecasted_data, historical_data_custom)
# fig = graphing.plot_historical_data(historical_data_plot, con_flag=True)
fig = graphing.plot_time_series_forecasts(historical_plot_df,
country_flag=True,
country_name=country)
except Exception as exc:
print("Error", exc)
else:
historical_data_custom = _historical_df.loc[_historical_df.index ==
country]
try:
forecasted_data = forecast_utils.get_forecasts(
country, constants.FORECAST_HORIZON)
historical_plot_df = data_utils.prep_plotting_data(
forecasted_data, historical_data_custom)
# fig = graphing.plot_historical_data(historical_data_plot)
fig = graphing.plot_time_series_forecasts(historical_plot_df,
country_flag=False,
country_name=country)
except Exception as exc:
print(exc)
historical_data = _historical_df.loc[_historical_df.index == country]
number_cases_confirmed = country_data["Confirmed"]
population = country_data["Population"]
num_hospital_beds = country_data["Num Hospital Beds"]
age_data = constants.AGE_DATA.loc[constants.AGE_DATA["State"] ==
country, :]
st.subheader(
f"How is the disease likely to spread in {country} in the next week?")
# Estimate true cases
true_cases_estimator = models.TrueInfectedCasesModel(
constants.ReportingRate.default)
# estimated_true_cases = true_cases_estimator.predict(number_cases_confirmed)
try:
week1_est = historical_plot_df.tail(1)
reported_vs_true_cases(
int(number_cases_confirmed), week1_est["confirmed"].tolist()[0],
graphing.abbreviate(week1_est["lower_bound"].tolist()[0],
round_factor=0),
graphing.abbreviate(week1_est["upper_bound"].tolist()[0],
round_factor=0))
# Plot historical data
st.write(fig)
print("Historical Data with Forecasts plotted")
except Exception as exc:
print(exc)
st.markdown(
f"Something went wrong :( Forecasts unavailable. Contact admin")
###################### SIR Model and Simulator ##############################
# Predict infection spread
sir_model = models.SIRModel(
transmission_rate_per_contact=constants.TransmissionRatePerContact.
default,
contact_rate=sidebar.contact_rate,
recovery_rate=constants.RecoveryRate.default,
normal_death_rate=constants.MortalityRate.default,
critical_death_rate=constants.CriticalDeathRate.default,
hospitalization_rate=constants.HospitalizationRate.default,
hospital_capacity=num_hospital_beds,
)
df = models.get_predictions(cases_estimator=true_cases_estimator,
sir_model=sir_model,
num_diagnosed=number_cases_confirmed,
num_recovered=country_data["Recovered"],
num_deaths=country_data["Deaths"],
area_population=population,
max_days=sidebar.num_days_for_prediction)
st.subheader("How will my actions affect the spread?")
st.write(
"**Use the slider in the sidebar to see how this changes the dynamics of disease spread**"
)
df_base = df[~df.Status.
isin(["Need Hospitalization", "Recovered", "Susceptible"])]
base_graph = graphing.infection_graph(df_base, df_base.Forecast.max(),
population * 0.5, population * 0.75)
# st.warning(graph_warning)
st.write(base_graph)
print("Infections Graph Plotted")
###################### Effect on hospitals ##############################
st.subheader("How will this affect my healthcare system?")
# Do some rounding to avoid beds sounding too precise!
approx_num_beds = round(num_hospital_beds / 100) * 100
st.write(
f"{country} has around **{approx_num_beds:,}** beds. Bear in mind that most of these "
"are probably already in use for people sick for other reasons.")
peak_occupancy = df.loc[df.Status ==
"Need Hospitalization"]["Forecast"].max()
percent_beds_at_peak = min(100 * num_hospital_beds / peak_occupancy, 100)
num_beds_comparison_chart = graphing.num_beds_occupancy_comparison_chart(
num_beds_available=approx_num_beds, max_num_beds_needed=peak_occupancy)
st.write(num_beds_comparison_chart)
st.markdown(
f"At peak, **{int(peak_occupancy):,}** people will need hospital beds. At least ** {100 - percent_beds_at_peak:.1f}% ** of "
f" people who need a bed in hospital might not have access given historical resources of {country}."
)
###################### Death Charts ##############################
st.subheader("How severe will the impact be?")
num_dead = df[df.Status == "Dead"].Forecast.iloc[-1]
num_recovered = df[df.Status == "Recovered"].Forecast.iloc[-1]
glob_hist = global_data.historical_country_data
uk_data = glob_hist.loc[(glob_hist.index == "UK") | \
(glob_hist.index == "United Kingdom"), :].copy()
uk_death_mirror = get_uk_death_mirror(uk_data, country_data["Deaths"])
death_plot = graphing.plot_death_timeseries(df[df.Status == "Dead"],
uk_death_mirror,
country_name=country)
st.markdown(
f"If the average person in {country} adopts the selected behavior, we estimate that **{int(num_dead):,}** "
f"people will die.")
st.markdown(f"This graph illustrates predicted deaths.")
outcomes_by_age_group = models.get_status_by_age_group(
num_dead, num_recovered, age_data)
fig = graphing.age_segregated_mortality(
outcomes_by_age_group.loc[:, ["Dead"]])
st.write(death_plot)
print("Deaths Graph Plotted")
st.markdown("-------")
st.markdown(body=generate_html("We'd love to hear from you!",
tag='h2',
color='#0090c4'),
unsafe_allow_html=True)
user_input = st.text_input("Enter your email to contact us")
utils.contact_us(user_input)
###################### Credits ##############################
st.subheader("References and Credits:")
st.markdown(
body=generate_html(
tag="h4",
text=
f"<u><a href=\"{NOTION_MODELLING_DOC}\" target=\"_blank\" style=color:{COLOR_MAP['pink']};>"
"Methodology</a></u> <span>  </span>"
"<hr>",
),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
tag="h4",
text=
f"<u><a href=\"https://github.com/CSSEGISandData/COVID-19\" target=\"_blank\" style=color:{COLOR_MAP['pink']};>"
"2019 Novel Coronavirus COVID-19 (2019-nCoV) Data Repository by Johns Hopkins CSSE</a></u> <span>  </span>"
"<hr>",
),
unsafe_allow_html=True,
)
print("complete....")
def run_app():
css.hide_menu()
css.limit_plot_size()
# Get cached country data
countries = _fetch_country_data()
if countries.stale:
st.caching.clear_cache()
countries = _fetch_country_data()
st.markdown(
body=generate_html(text=f"Corona Calculator", bold=True, tag="h1"),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
tag="h2",
text=
"A tool to help you visualize the impact of social distancing <br>",
),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
text=
"<strong>Disclaimer:</strong> <em>The creators of this application are not healthcare professionals. "
"The illustrations provided were estimated using best available data but might not accurately reflect reality.</em>",
color="gray",
font_size="12px",
),
unsafe_allow_html=True,
)
st.markdown(
body=generate_html(
tag="h4",
text=
f"<u><a href=\"{NOTION_MODELLING_DOC}\" target=\"_blank\" style=color:{COLOR_MAP['pink']};>"
"Methodology</a></u> <span>  </span>"
f"<u><a href=\"{MEDIUM_BLOGPOST}\" target=\"_blank\" style=color:{COLOR_MAP['pink']};>"
"Blogpost</a> </u>"
"<hr>",
),
unsafe_allow_html=True,
)
sidebar = Sidebar(countries)
country = sidebar.country
country_data = countries.country_data[country]
number_cases_confirmed = country_data["Confirmed"]
population = country_data["Population"]
num_hospital_beds = country_data["Num Hospital Beds"]
sir_model = models.SIRModel(
transmission_rate_per_contact=constants.TransmissionRatePerContact.
default,
contact_rate=sidebar.contact_rate,
recovery_rate=constants.RecoveryRate.default,
normal_death_rate=constants.MortalityRate.default,
critical_death_rate=sidebar.severe_mortality_rate,
hospitalization_rate=constants.HospitalizationRate.default,
hospital_capacity=num_hospital_beds,
)
true_cases_estimator = models.TrueInfectedCasesModel(
constants.AscertainmentRate.default)
df = models.get_predictions(
true_cases_estimator,
sir_model,
number_cases_confirmed,
population,
sidebar.num_days_for_prediction,
)
reported_vs_true_cases(
number_cases_confirmed,
true_cases_estimator.predict(number_cases_confirmed))
st.write(
"The number of reported cases radically underestimates the true cases, because people do not show symptoms for "
"several days, not everybody gets tested, and the tests take a few days to return results. "
"The extent depends upon your country's testing strategy."
" We estimated the above using numbers from Japan ([source](https://www.ncbi.nlm.nih.gov/pubmed/32033064))."
)
st.subheader("How will the disease spread?")
st.write(
"The critical factor for controlling spread is how many others infected people interact with each day. "
"This has a dramatic effect upon the dynamics of the disease. ")
st.write(
"**Play with the slider to the left to see how this changes the dynamics of disease spread**"
)
df_base = df[~df.Status.isin(["Need Hospitalization", "Need Ventilation"])]
base_graph = graphing.infection_graph(df_base, df.Forecast.max())
st.warning(graph_warning)
st.write(base_graph)
hospital_graph = graphing.hospitalization_graph(
df[df.Status.isin(["Infected", "Need Hospitalization"])],
num_hospital_beds,
max(num_hospital_beds, df.Forecast.max()),
)
st.write(
"Note that we use a fixed estimate of the mortality rate here, of 1% [(source)](https://institutefordiseasemodeling.github.io/nCoV-public/analyses/first_adjusted_mortality_estimates_and_risk_assessment/2019-nCoV-preliminary_age_and_time_adjusted_mortality_rates_and_pandemic_risk_assessment.html). "
"In reality, the mortality rate will be highly dependent upon the load upon the healthcare system and "
"the availability of treatment. Some estimates ([like this one](https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30195-X/fulltext)) are closer to 6%."
)
st.subheader("How will this affect my healthcare system?")
st.write(
"The important variable for hospitals is the peak number of people who require hospitalization"
" and ventilation at any one time.")
# Do some rounding to avoid beds sounding too precise!
st.write(
f"Your country has around **{round(num_hospital_beds / 100) * 100:,}** beds. Bear in mind that most of these "
"are probably already in use for people sick for other reasons.")
st.write(
"It's hard to know how many ventilators are present per country, but there will certainly be a worldwide "
"shortage. Many countries are scrambling to buy them [(source)](https://www.reuters.com/article/us-health-coronavirus-draegerwerk-ventil/germany-italy-rush-to-buy-life-saving-ventilators-as-manufacturers-warn-of-shortages-idUSKBN210362)."
)
st.warning(graph_warning)
st.write(hospital_graph)
peak_occupancy = df.loc[df.Status ==
"Need Hospitalization"]["Forecast"].max()
percent_beds_at_peak = min(100 * num_hospital_beds / peak_occupancy, 100)
st.markdown(
f"At peak, **{peak_occupancy:,}** people will need hospital beds. ** {percent_beds_at_peak:.1f} % ** of people "
f"who need a bed in hospital will have access to one given your country's historical resources. This does "
f"not take into account any special measures that may have been taken in the last few months."
)
def train(mhc, data, model, model_path, lr, n_epoch, transfer_path):
'''
Training protocol
'''
# print out options
print(
'Training\nMHC: %s\nData: %s\nModel: %s\nSave path: %s\nTransfer: %s' %
(mhc, data, model, model_path, transfer_path))
# load training
train_data = Dataset.from_csv(filename=data,
sep=',',
allele_column_name='mhc',
peptide_column_name='peptide',
affinity_column_name='IC50(nM)')
# apply cut/pad or mask to same length
if 'lstm' in model or 'gru' in model:
train_data.mask_peptides()
else:
train_data.cut_pad_peptides()
# get the allele specific data
mhc_train = train_data.get_allele(mhc)
# define model
if model == 'fc':
model = models.mhcnuggets_fc()
elif model == 'gru':
model = models.mhcnuggets_gru()
elif model == 'lstm':
model = models.mhcnuggets_lstm()
elif model == 'chunky_cnn':
model = models.mhcnuggets_chunky_cnn()
elif model == 'spanny_cnn':
model = models.mhcnuggets_spanny_cnn()
# check if we need to do transfer learning
if transfer_path:
model.load_weights(transfer_path)
# compile model
model.compile(loss='mse', optimizer=Adam(lr=0.001))
# get tensorized values for training
train_peptides, train_continuous, train_binary = mhc_train.tensorize_keras(
embed_type='softhot')
# convergence criterion
highest_f1 = -1
for epoch in range(n_epoch):
# train
model.fit(train_peptides, train_continuous, epochs=1, verbose=0)
# test model on training data
train_preds_cont, train_preds_bin = get_predictions(
train_peptides, model)
train_auc = roc_auc_score(train_binary, train_preds_cont)
train_f1 = f1_score(train_binary, train_preds_bin)
train_ktau = kendalltau(train_continuous, train_preds_cont)[0]
print('epoch %d / %d' % (epoch, n_epoch))
print('Train AUC: %.4f, F1: %.4f, KTAU: %.4f' %
(train_auc, train_f1, train_ktau))
# convergence
if train_f1 > highest_f1:
highest_f1 = train_f1
best_epoch = epoch
model.save_weights(model_path)
print('Done!')
# -*- coding: utf-8 -*-
"""
Main file for the second Machine Learning project:
### TODO ###
Authors: Kirill IVANOV, Matthias RAMIREZ, Nicolas TALABOT
"""
import argparse
from models import get_pretrained_models, train_and_get_models, get_predictions
# Argument parser
parser = argparse.ArgumentParser()
parser.add_argument(
"--train_model",
help="train the whole model, do not reuse the pre-trained one",
action="store_true")
args = parser.parse_args()
data_dir = 'training/'
test_data_dir = 'test_set_images/'
model1, model2, model3 = train_and_get_models(
data_dir) if args.train_model else get_pretrained_models()
get_predictions(model1, model2, model3, test_data_dir)
