def plot_moreprojects():
data = read_csv("moreprojects")
data = data[(data.hungry <= 40) & data.success]
# drop known outliers:
data = data[~((data.projects == 6) & (data.hungry == 35))]
data = data[~((data.projects == 8) & (data.hungry == 39))]
projects = [5, 6, 7, 8, 9]
labels = [f"{n} projects" for n in projects]
x_ticks = data[data.projects == projects[0]].hungry.unique()
fig, ax = resultlib.prepare_graph()
for n, label in zip(projects, labels):
subsel = data[data.projects == n][["hungry", "nsec"]]
grouping = subsel.groupby("hungry")["nsec"]
line = grouping.median()
ax.plot(line.index, line.values, "-", label=label)
ax.set_xlabel("Number of workers")
plt.legend()
plt.xticks(x_ticks, rotation="vertical")
# plt.yscale("log")
fig.tight_layout()
plt.savefig("workers-moreprojects.pdf")
plt.close()
def plot_morerooms():
data = read_csv("morerooms-optimizing")
data = data[["hungry", "nsec"]][data.hungry < 16]
grouping = data.groupby("hungry")["nsec"]
xticks, series = zip(*grouping)
fig, ax = resultlib.prepare_graph()
ax.boxplot(series, showfliers=False, positions=xticks)
# plt.xticks(range(1, len(series) + 1), xticks)
def exp(x, a, b, c):
return a * np.exp(b * x) + c
meds = grouping.median()
popt, _ = curve_fit(exp, meds.index, meds.values)
fitx = np.linspace(xticks[0], xticks[-1], 100)
fity = exp(fitx, *popt)
plt.plot(fitx, fity, "-")
ax.set_xlabel("Number of workers")
fig.tight_layout()
plt.savefig("workers-morerooms.pdf")
plt.close()
def plot_badsolver():
data = read_csv("badsolver-growingprojects")
groups = data.groupby("projects")["nsec"]
projects = list(range(4, 31, 4))
fig, ax = resultlib.prepare_graph()
for p in projects:
series = groups.get_group(p)
series = series.sort_values(ascending=True).reset_index(drop=True)
ax.plot(series.index, series, "-", label=f"{p} projects")
perc_ticks = list(range(0, 101, 10))
plt.xticks([p * 5 for p in perc_ticks], perc_ticks)
ax.set_xlabel("Percentile")
plt.legend()
fig.tight_layout()
plt.savefig("badsolver.pdf")
plt.close()
def plot_timeouts():
data = resultlib.read_csv("timelimits", COLUMNS)
data.timeout //= 1000
flt = data[data.success]
grouping = flt.groupby("timeout")["utility"]
xticks, series = zip(*grouping)
fig, ax = resultlib.prepare_graph()
plt.xticks(rotation="vertical")
ax.boxplot(series, showfliers=False, positions=xticks)
def log(x, a, b):
return a + b * np.log(x)
meds = grouping.median()
popt, _ = curve_fit(log, meds.index, meds.values)
fitx = np.linspace(xticks[0], xticks[-1], 100)
fity = log(fitx, *popt)
ax.plot(fitx, fity, "-")
color = "#d62728"
ax2 = ax.twinx()
ax2.set_yticks(list(range(0, 101, 10)))
ax2.set_ylim(0, 100)
perc = 100 - grouping.count()
ax2.plot(perc.index, perc.values, "--", color=color)
ax2.set_ylabel("Failed attempts (out of 100)")
ax2.tick_params(axis="y", labelcolor=color)
plt.xlim(0.5, 30.5)
ax.set_xlabel("Time limit (s)")
ax.set_ylabel("Total utility")
fig.tight_layout()
plt.savefig("timelimits.pdf")
plt.close()
def plot_onebyone():
data = read_csv("oneworker-params")
data = data[data.success]
x_ticks = data.lunch_n.unique()
project_counts = data.projects.unique()
colors = make_colors(len(project_counts))
labels = [f"{n} projects" for n in project_counts]
fig, ax = resultlib.prepare_graph()
def poly(x, a, b, e):
return a * x ** e + b
for label, color, n in zip(labels, colors, project_counts):
subsel = data[data.projects == n][["lunch_n", "nsec"]]
grouping = subsel.groupby("lunch_n")["nsec"]
linedata = grouping.mean()
lineerr = grouping.std()
ax.errorbar(
linedata.index,
linedata.values,
yerr=lineerr,
fmt="o",
label=label,
color=color,
)
popt, _ = curve_fit(poly, linedata.index, linedata.values)
fitx = np.linspace(x_ticks[0], x_ticks[-1], 100)
fity = poly(fitx, *popt)
plt.plot(fitx, fity, "-", color=colors[1])
plt.xticks(x_ticks, rotation="45")
plt.legend(loc="upper left")
ax.set_xlabel("Number of lunch rooms")
fig.tight_layout()
plt.savefig("workers-oneworker.pdf")
plt.close()
def plot_variables():
fig, ax = resultlib.prepare_graph()
for ifn, header, fmt in zip(FILES, HEADERS, FORMATS):
data = read_csv(ifn)
subsel = data[data.success]
subsel = subsel[[ifn, "nsec"]]
if ifn == "constraints":
subsel.constraints /= 100
grouping = subsel.groupby(ifn)["nsec"]
line = grouping.mean()
ax.plot(line.index, line.values, fmt, label=header)
x_ticks = range(500, 10001, 500)
plt.xticks(x_ticks, rotation="vertical")
# fig, ax = box_graph(datasets, labels, colors, x_ticks, "vertical")
ax.set_xlabel("Number of elements")
plt.legend()
fig.tight_layout()
plt.savefig("variables.pdf")
plt.close()
def plot_simple():
data = read_csv("workercount-simple")
projects = data.projects.unique()
fmts = {5: "-o", 15: "-v", 50: "-x"}
fig, ax = resultlib.prepare_graph()
for n in projects:
subsel = data[data.projects == n][["workers", "nsec"]]
grouping = subsel.groupby("workers")["nsec"]
line = grouping.median()
lineerr = grouping.std()
ax.errorbar(
line.index, line.values, yerr=lineerr, fmt=fmts[n], label=f"{n} projects"
)
plt.legend()
plt.xticks(line.index, rotation="45")
ax.set_xlabel("Number of workers")
fig.tight_layout()
plt.savefig("workers-simple.pdf")
plt.close()
def plot_variablemem():
fig, ax = resultlib.prepare_graph()
for ifn, header, fmt in zip(FILES, HEADERS, FORMATS):
data = read_csv(ifn)
data.memory *= 1e-6
subsel = data[data.success]
subsel = subsel[[ifn, "memory"]]
if ifn == "constraints":
subsel.constraints /= 100
grouping = subsel.groupby(ifn)["memory"]
line = grouping.max()
ax.plot(line.index, line.values, fmt, label=header)
x_ticks = range(500, 10001, 500)
plt.xticks(x_ticks, rotation="vertical")
ax.set_ylabel("Peak memory usage (MB)")
ax.set_xlabel("Number of elements")
plt.legend()
fig.tight_layout()
plt.savefig("variablemem.pdf")
plt.close()