def plot_fitness_sim_results(df, params,
ymin=None,
ymax=None,
yticks=None,
title=None,
x_step=10):
sns.set_style("ticks")
# plot population size
popsizes = fitness.str_popsizes_to_array(df["log_pop_sizes"])
# take the total population size: sum of populations tuned to any
# nutrient state
df["log2_pop_size"] = np.log2(np.exp(popsizes).sum(axis=1))
policy_colors = params["policy_colors"]
# only plot policies we have colors for
policies_with_colors = policy_colors.keys()
df = df[df["policy"].isin(policies_with_colors)]
init_pop_size = sum(params["init_pop_sizes"])
# group results by time and by policy
grouped = df.groupby(["t", "policy"], as_index=False)
summary_df = grouped.agg({"log2_pop_size": [np.mean, np.std],
"growth_rates": [np.mean, np.std]})
rand_summary = summary_df[summary_df["policy"] == "Random"]
time_obj = time_unit.Time(params["t_start"],
params["t_end"],
step_size=params["step_size"])
step_size = params["step_size"]
final_df = summary_df[summary_df["t"] == time_obj.t[-1]]
## plot population size across time
def plot_pop_size_across_time(params,
ymin=ymin,
ymax=ymax):
offset = step_size / 250.
num_xticks = 11
ax = sns.tsplot(time="t", value="log2_pop_size", unit="sim_num",
condition="policy", color=policy_colors,
err_style="ci_band",
ci=95,
data=df,
legend=False)
for policy_num, policy in enumerate(policy_colors):
error_df = summary_df[summary_df["policy"] == policy]
c = policy_colors[policy]
assert (len(error_df["t"]) == len(time_obj.t) == \
len(error_df["log2_pop_size"]["mean"]))
plt.xlabel("Time step", fontsize=10)
plt.ylabel("Pop. size ($\log_{2}$)", fontsize=10)
# assuming glucose is listed first
gluc_growth_rate = params["nutr_growth_rates"][0]
galac_growth_rate = params["nutr_growth_rates"][1]
if title is not None:
plt.title(title, fontsize=8)
else:
plt.title(r"$P_{0} = %d$, " \
r"$\mu_{\mathsf{Glu}} = %.2f, \mu_{\mathsf{Gal}} = %.2f$, " \
r"$\mu_{\mathsf{Mis}} = %.2f$, lag = %d, " \
r"%d iters" %(init_pop_size,
gluc_growth_rate,
galac_growth_rate,
params["mismatch_growth_rate"],
params["decision_lag_time"],
params["num_sim_iters"]),
fontsize=8)
c = 0.5
plt.xlim([min(df["t"]) - c, max(df["t"]) + c])
if ymin is None:
ymin = int(np.log2(init_pop_size))
plt.ylim(ymin=ymin)
plt.xlim([time_obj.t.min(),
time_obj.t.max()])
plt.xticks(range(int(time_obj.t.min()), int(time_obj.t.max()) + x_step,
x_step),
fontsize=8)
if yticks is not None:
plt.yticks(yticks, fontsize=8)
plt.ylim(yticks[0], yticks[-1])
sns.despine(trim=True, offset=2*time_obj.step_size)
plt.tick_params(axis='both', which='major', labelsize=8,
pad=2)
# make plot
plot_pop_size_across_time(params)
Get spline fit.
"""
spl = scipy.interpolate.UnivariateSpline(x, y, k=4, s=1)
return spl(x)
def get_spline_derivs_fit(x, y):
"""
Get spline derivatives fit.
"""
spl = scipy.interpolate.UnivariateSpline(x, y, k=4, s=1)
derivs = spl.derivative()
return derivs(x)
sim_to_plot = sims_to_plot[0]
df = results[sim_to_plot]["data"]
popsizes = fitness.str_popsizes_to_array(df["log_pop_sizes"])
# take the total population size: sum of populations tuned to any
# nutrient state
df["log2_pop_size"] = np.log2(np.exp(popsizes).sum(axis=1))
print df
g = df.groupby(["policy", "sim_num"])
new_df = []
for name, group in g:
curr_df = group.copy()
print "fitting to: ", curr_df["t"], curr_df["growth_rates"]
curr_df["fitted_growth_rate"] = \
get_spline_fit(curr_df["t"], curr_df["growth_rates"])
curr_df["fitted_growth_rate_from_pop"] = \
get_spline_derivs_fit(curr_df["t"], curr_df["log2_pop_size"])
plt.figure()
plt.plot(curr_df["t"], curr_df["growth_rates"])