def visualise_tsne(tsne_df, save=True, fig_num=15):
"""Visualise tsne plot"""
tsne_base = alt.Chart(tsne_df).encode(
x=alt.X("x:Q", title="", axis=alt.Axis(ticks=False, labels=False)),
y=alt.Y("y:Q", title="", axis=alt.Axis(ticks=False, labels=False)),
)
tsne_points = ((
tsne_base.mark_point(
filled=True, opacity=0.5, stroke="black",
strokeOpacity=0.5).encode(
color=alt.Color("org_type", title="Organisation type"),
strokeWidth=alt.Stroke("top",
scale=alt.Scale(range=[0, 1]),
legend=None),
# stroke = alt.value('blue'),
size=alt.Size("activity:Q", title="Number of papers"),
facet=alt.Facet("size",
columns=2,
title="Number of organisations in plot"),
tooltip=["index"],
)).interactive().resolve_scale(
y="independent", x="independent").properties(width=250,
height=250))
if save is True:
save_altair(tsne_points, "fig_15_tsne", driv)
return tsne_points
def main():
st.title('Covid-19 SIR Model')
days=st.slider('Select no. of days :',0,100,1,1)
Istart=st.slider('Intially infected pop % :',.01,1.0,.01,.01)
trans_rate=st.slider('Transfer rate :',0,10,1,1)
reco_rate=st.slider('Recovery rate :',0.0,1.0,0.1,0.01)
df = corona(Istart,trans_rate,reco_rate,100,days)
df_melt=pd.melt(df,id_vars='Days',value_vars=['Suspect','Infected','Recovered'],var_name='Legend',value_name='% of Population')
st.altair_chart(alt.layer(
Charts(df_melt,'Infected','darkred'),
Charts(df_melt,'Suspect','darkblue'),
Charts(df_melt,'Recovered','darkgreen')).encode(
alt.Stroke(
'clm:N',scale=alt.Scale(domainMid=1,domain=['Infected','Suspect','Recoverd'],
range=['darkred','darkblue','darkgreen'] ))))
def test_convert_stroke_fail_temporal(column):
chart = alt.Chart(df).mark_point().encode(alt.Stroke(column))
convert(chart)
def app():
# Load 6,000 rows of data into the dataframe.
bData = load_data(bFile)
rRawData = load_data(rFile, rDateCol, 6000)
# add col to bData for int version of ratings
bData['starsInt'] = [int(i) for i in bData['stars']]
# parse rData to add year column
rRawData['year'] = pd.DatetimeIndex(rRawData['date'], tz='UTC').year
rData = copy.copy(rRawData)
rData = rData.drop(columns=['date'])
# st.subheader('Dropped date col')
# st.write(rData.head())
yearSet = set(rData['year'])
bSmallData = bData.head()
rSmallData = rData.head()
#uSmallData = uData.head()
################################################################################
# Titles and Information
################################################################################
st.title('Explore City-Wide Restaurant Reviews')
st.subheader(
'Check out the ratings of restaurants in Pittsburgh and other major cities, their reviews, and their sentiment over time.'
)
################################################################################
# Maps
################################################################################
# MAP VISUALIZATION
# break apart data into city specific data
pittsburghData = bData.loc[bData['city'] == 'Pittsburgh']
montrealData = bData.loc[bData['city'] == 'Montréal']
clevelandData = bData.loc[bData['city'] == 'Cleveland']
torontoData = bData.loc[bData['city'] == 'Toronto']
st.header('City Restaurant Data')
# slider and data from user input
#st.map(bData)
selectedStars = st.sidebar.slider('Filter maps by selecting ratings:', 0,
5, (0, 5))
filteredDataP = pittsburghData[
(pittsburghData['starsInt'] >= selectedStars[0])
& (pittsburghData['starsInt'] <= selectedStars[1])]
filteredDataM = montrealData[
(montrealData['starsInt'] >= selectedStars[0])
& (montrealData['starsInt'] <= selectedStars[1])]
filteredDataC = clevelandData[
(clevelandData['starsInt'] >= selectedStars[0])
& (clevelandData['starsInt'] <= selectedStars[1])]
filteredDataT = torontoData[(torontoData['starsInt'] >= selectedStars[0]) &
(torontoData['starsInt'] <= selectedStars[1])]
# PREFERENCE OF FILTERED DATA
# initiate bar chart below maps
starsBarP = barChart = alt.Chart(
pittsburghData,
title='Pittsburgh Restaurant Rating').mark_bar().encode(
alt.X('count():Q', title='Number of reviews'),
alt.Y('starsInt:N', title='Stars Rated'),
alt.Color('starsInt:Q', legend=None))
starsBarM = barChart = alt.Chart(
montrealData, title='Montreal Restaurant Rating').mark_bar().encode(
alt.X('count():Q', title='Number of reviews'),
alt.Y('starsInt:N', title='Stars Rated'),
alt.Color('starsInt:Q', legend=None))
starsBarC = barChart = alt.Chart(
clevelandData, title='Cleveland Restaurant Rating').mark_bar().encode(
alt.X('count():Q', title='Number of reviews'),
alt.Y('starsInt:N', title='Stars Rated'),
alt.Color('starsInt:Q', legend=None))
starsBarT = barChart = alt.Chart(
torontoData, title='Toronto Restaurant Rating').mark_bar().encode(
alt.X('count():Q', title='Number of reviews'),
alt.Y('starsInt:N', title='Stars Rated'),
alt.Color('starsInt:Q', legend=None))
#CONFIGURE LAYOUT
# first 2 maps
col1, col2 = st.beta_columns(2)
col1.subheader('Pittsburgh')
col1.map(filteredDataP)
col1.write(starsBarP)
col2.subheader('Montreal')
col2.map(filteredDataM)
col2.write(starsBarM)
# second 2 maps
col3, col4 = st.beta_columns(2)
col3.subheader('Cleveland')
col3.map(filteredDataC)
col3.write(starsBarC)
col4.subheader('Toronto')
col4.map(filteredDataT)
col4.write(starsBarT)
################################################################################
# Exploratory Visuals
################################################################################
# PREFERENCE OVER TIME
rDataFiltered = copy.copy(rData)
cola, colb, colc = st.beta_columns(3)
pref = 'neu'
prefTitle = 'Neutral '
st.subheader('View colorized sentiment overtime.')
st.write('Select review type to filter preference overtime')
if cola.button('Negative Reviews'):
#rDataFiltered = rDataFiltered[rDataFiltered['neg']]
pref = 'neg'
prefTitle = 'Negative '
if colb.button('Positive Reviews'):
#rDataFiltered = rDataFiltered[rDataFiltered['pos']]
pref = 'pos'
prefTitle = 'Positive '
if colc.button('Neutral Reviews'):
#rDataFiltered = rDataFiltered[rDataFiltered['neu']]
pref = 'neu'
prefTitle = 'Neutral '
st.write('Preference Over Time:')
preference = alt.Chart(
rDataFiltered,
title='Preference Over Time').mark_circle(strokeWidth=2).encode(
alt.X('year:N', title='Date', axis=alt.Axis(labelAngle=0)),
alt.Y('bcity:N', title='City'),
alt.Size('count():Q',
scale=alt.Scale(range=[0, 4000]),
legend=alt.Legend(title='Number of Reviews')),
#alt.Opacity(pref, type = 'quantitative', aggregate = 'mean'),
alt.Stroke('bcity:N', legend=None),
alt.Color(pref, type='quantitative', aggregate='mean'),
alt.Tooltip('neg:Q', aggregate='mean')).properties(height=440,
width=800)
# map total average pref reviews over time
totalPreference = alt.Chart(
rDataFiltered, title=prefTitle + 'Over Time').mark_area(
opacity=0.8, color='black').encode(
alt.X('year:O', axis=alt.Axis(labelAngle=0), title='Year'),
alt.Y(pref,
type='quantitative',
title=prefTitle + 'Sentiment Average')).properties(
height=150, width=800)
st.write(preference & totalPreference)
def test_quantitative_stroke():
chart = alt.Chart(df_quant).mark_point().encode(alt.Stroke('fill'))
convert(chart)
def test_line_stroke(self, x, y, s):
chart = alt.Chart(df_line).mark_line().encode(alt.X(x), alt.Y(y),
alt.Stroke(s))
fig, _ = convert(chart)
return fig
def noise_plot(source=data['students']['kd']['noise'],
title="Diferencia con ResNet101 contra ruido",
y='delta_teacher',
x='noise',
color='feat_dist',
shape='layer',
student='ResNet18',
xscale='linear',
yscale='linear'):
d = locals()
ks = [
i for i in d.keys()
if i not in ['source', 'title', 'xscale', 'yscale', 'scale', 'bs']
]
vals = [d[i] for i in ks]
source = source.drop(columns=[
i for i in source.columns if i not in vals + ['student', 'noise']
])
sou = source[source['student'] == student]
#reduce data
selection = alt.selection_multi(fields=[color], bind='legend')
ytitle = global_titles[
y] if yscale == 'linear' else global_titles[y] + " [%s]" % yscale
xtitle = global_titles[
x] if xscale == 'linear' else global_titles[x] + " [%s]" % xscale
base = alt.Chart(sou, title=title).mark_line().encode(
alt.Y(y,
type='quantitative',
title=ytitle,
scale=alt.Scale(zero=False, base=10, type=yscale)),
alt.X('noise:Q',
type='quantitative',
scale=alt.Scale(zero=False, base=10, type=xscale),
title=xtitle),
color=alt.Color('%s:N' % color,
legend=alt.Legend(title=global_titles[color]),
scale=alt.Scale(scheme='spectral')),
strokeDash=alt.Stroke('%s:N' % shape,
legend=alt.Legend(title=global_titles[shape])),
opacity=alt.condition(
selection, alt.value(1),
alt.value(0.2))).add_selection(selection).properties(width=600,
height=600)
line = alt.Chart(sou).mark_line().encode(
x=x,
#y='mean(%s)'%y,
y=alt.Y('mean(%s)' % y,
type='quantitative',
title=ytitle,
scale=alt.Scale(zero=False, base=10, type=yscale)),
color=alt.Color('%s:N' % color,
legend=alt.Legend(title=global_titles[color]),
scale=alt.Scale(scheme='spectral')),
opacity=alt.condition(selection, alt.value(1), alt.value(0.2)),
shape=alt.Shape('%s:N' % color, legend=alt.Legend(title='media')),
#opacity=alt.value(0.5)
)
point = alt.Chart(sou).mark_point(filled=True).encode(
x=x,
#y='mean(%s)'%y,
y=alt.Y('mean(%s)' % y,
type='quantitative',
title=ytitle,
scale=alt.Scale(zero=False, base=10, type=yscale)),
color=alt.Color('%s:N' % color,
legend=alt.Legend(title=global_titles[color]),
scale=alt.Scale(scheme='spectral')),
opacity=alt.condition(selection, alt.value(1), alt.value(0.2)),
shape=alt.Shape('%s:N' % color, legend=alt.Legend(title='media')),
#opacity=alt.value(0.5)
)
band = alt.Chart(sou).mark_errorband(extent='ci').encode(
x=alt.X('noise', title=xtitle),
y=alt.Y('%s' % y, title=global_titles[y]),
color=alt.Color('%s:N' % color,
legend=alt.Legend(title=global_titles[color]),
scale=alt.Scale(scheme='spectral')),
opacity=alt.condition(selection, alt.value(0.4), alt.value(0.05)))
w = base + band + line + point
w.properties(width=600, height=600)
return w
def grafico_con_barra_y_CE(
source=data['students']['kd']['KD'],
title="Diferencia de accuracy en entrenamiento con respecto a Resnet101.",
y="delta_teacher",
x="feat_dist",
fill='layer',
color='student',
shape='layer',
bs=True,
scale='linear'):
if x == 'dists':
source = source.copy()
source['dists'] = source['feat_dist'] + ", " + source['log_dist']
elif x == 'feat,block':
source = source.copy()
source['feat,block'] = [
i + ", " + str(j)
for i, j in list(zip(source['feat_dist'], source['layer']))
]
#reduce data
d = locals()
ks = [
i for i in d.keys()
if i not in ['source', 'title', 'xscale', 'yscale', 'scale', 'bs']
]
vals = [d[i] for i in ks]
source = source.drop(columns=[i for i in source.columns if i not in vals])
chart = alt.Chart(source, title=title).mark_point(size=100).encode(
y=alt.Y(
y,
type='quantitative',
scale=alt.Scale(zero=True, base=2, constant=1, type=scale),
title=global_titles[y] if scale == 'linear' else global_titles[y] +
" [%s]" % scale),
x=alt.X(x, title=global_titles[x]),
color=alt.Color(color, legend=alt.Legend(title=global_titles[color])),
fill=alt.Fill('%s:O' % fill,
legend=alt.Legend(title=global_titles[fill]),
scale=alt.Scale(scheme='pastel1')),
shape=alt.Shape("%s:O" % shape,
legend=alt.Legend(title=global_titles[shape])),
opacity=alt.value(0.5))
bar = alt.Chart(bar_source).mark_rule(opacity=0.5).encode(
y=y,
color=alt.Color('model', legend=alt.Legend(title="Modelo")),
stroke=alt.Stroke('model',
legend=alt.Legend(title="Modelo en Cross Entropy")),
size=alt.value(2))
if bs:
bar = alt.Chart(bar_source).mark_rule(opacity=0.5).encode(
y=y,
color=alt.Color('model', legend=alt.Legend(title="Modelo")),
stroke=alt.Stroke(
'model', legend=alt.Legend(title="Modelo en Cross Entropy")),
size=alt.value(2))
d = bar + chart
d.properties(width=600, height=600).configure_axis(
titleFontSize=12).configure_title(fontSize=15)
return d
chart.properties(width=600, height=600).configure_axis(
titleFontSize=12).configure_title(fontSize=15)
return chart
def grafico_con_barra_xy_CE(
source=data['students']['kd']['KD'],
title="Ratio entre exactitud en validacion y accuracy en entrenamiento.",
y="test/train",
x="temp",
color='student',
shape='log_dist',
bs=True,
fill=None,
xscale='log',
yscale='linear'):
#reduce data
d = locals()
ks = [
i for i in d.keys()
if i not in ['source', 'title', 'xscale', 'yscale', 'scale', 'bs']
]
vals = [d[i] for i in ks]
source = source.drop(columns=[i for i in source.columns if i not in vals])
ytitle = global_titles[
y] if yscale == 'linear' else global_titles[y] + " [%s]" % yscale
xtitle = global_titles[
x] if xscale == 'linear' else global_titles[x] + " [%s]" % xscale
encodings = {
'y':
alt.Y(y,
type='quantitative',
title=ytitle,
scale=alt.Scale(zero=False, base=10, type=yscale)),
'x':
alt.X(x,
type='quantitative',
title=xtitle,
scale=alt.Scale(zero=False, base=10, type=xscale)),
'color':
alt.Color(color, legend=alt.Legend(title=global_titles[color])),
'shape':
alt.Shape(shape, legend=alt.Legend(title=global_titles[shape])),
#size=50
}
if fill is not None:
encodings["fill"] = alt.Fill(fill)
chart = alt.Chart(
source,
title=title).mark_point(size=100).encode(**encodings).interactive()
if bs:
bar = alt.Chart(bar_source).mark_rule(opacity=0.5).encode(
y=y,
color=alt.Color('model', legend=alt.Legend(title="Modelo")),
stroke=alt.Stroke(
'model', legend=alt.Legend(title="Modelo en Cross Entropy")),
size=alt.value(2))
d = bar + chart
d.properties(width=600, height=600).configure_axis(
titleFontSize=12).configure_title(fontSize=15)
return d
chart.properties(width=600, height=600).configure_axis(
titleFontSize=12).configure_title(fontSize=15)
return chart
if colc.button('Neutral Reviews'):
#rDataFiltered = rDataFiltered[rDataFiltered['neu']]
pref = 'neu'
prefTitle = 'Neutral '
st.write('Preference Over Time:')
methods2 = alt.Chart(
rDataFiltered,
title='Preference Over Time').mark_circle(strokeWidth=2).encode(
alt.X('year:N', title='Date', axis=alt.Axis(labelAngle=0)),
alt.Y('bcity:N', title='City'),
alt.Size('count():Q',
scale=alt.Scale(range=[0, 4000]),
legend=alt.Legend(title='Number of Reviews')),
#alt.Opacity(pref, type = 'quantitative', aggregate = 'mean'),
alt.Stroke('bcity:N', legend=None),
alt.Color(pref, type='quantitative', aggregate='mean'),
alt.Tooltip('neg:Q', aggregate='mean')).properties(height=440,
width=800)
#st.write(methods2)
# map total average pref reviews over time
total = alt.Chart(rDataFiltered, title=prefTitle + 'Over Time').mark_area(
opacity=0.8, color='black').encode(
alt.X('year:O', axis=alt.Axis(labelAngle=0), title='Year'),
alt.Y(pref,
type='quantitative',
title=prefTitle + 'Average Preference')).properties(height=150,
width=800)
st.write(methods2 & total)