Skip to content

Commit

Permalink
[analysis] add with resulting plots
Browse files Browse the repository at this point in the history
  • Loading branch information
@Stunkymonkey
Stunkymonkey committed Sep 15, 2021
1 parent 8178790 commit 481a395
Show file tree
Hide file tree
Showing 10 changed files with 1,454 additions and 206 deletions.
210 changes: 176 additions & 34 deletions analysis/graph-details.ipynb
View file
Original file line number Diff line number Diff line change
Expand Up @@ -35,12 +35,13 @@
"metadata": {},
"outputs": [],
"source": [
"EVAL_DIR = \"/home/felix/todo/osm-tmp\"\n",
"OUTPUT = \"single\"\n",
"EVAL_DIR = \"/home/felix/todo/algohol/single\"\n",
"MLP_METHODS = [\"kmeans\", \"gonzalez\", \"merge\"]\n",
"MLP_LEVELS = [[int(2 ** i)] for i in np.arange(8.0, 13.5, 1.0)]\n",
"MLP_LEVELS = [[int(2 ** i)] for i in np.arange(9.0, 12.5, 1.0)]\n",
"FAST_QUERY_METHODS = [\"pcrp\", \"pch\", \"prp\"]\n",
"QUERY_METHODS = [\"normal\", \"bi\"] + FAST_QUERY_METHODS\n",
"AREAS = [\"saarland\"]\n",
"QUERY_METHODS = [\"normal\"] + FAST_QUERY_METHODS\n",
"AREAS = [\"baden-wuerttemberg\"]\n",
"print(MLP_METHODS, \"with\", MLP_LEVELS)"
]
},
Expand Down Expand Up @@ -80,46 +81,41 @@
"metadata": {},
"outputs": [],
"source": [
"assert((df_graph.groupby([\"amount_edges\"]).size() == 5).all())"
"assert((df_graph.groupby([\"amount_edges\"]).size() == len(QUERY_METHODS)).all())"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "amber-cycle",
"id": "unauthorized-contest",
"metadata": {},
"outputs": [],
"source": [
"fig, ax = plt.subplots()\n",
"speedups = list()\n",
"for area in AREAS:\n",
" dijkstra = df_graph[(df_graph.Query == \"normal\") & (df_graph.Area == area)][\"amount_used_edges\"].iloc[0]\n",
" bidijkstra = df_graph[(df_graph.Query == \"bi\") & (df_graph.Area == area)][\"amount_used_edges\"].iloc[0]\n",
" for query in FAST_QUERY_METHODS:\n",
" for mlp in MLP_METHODS:\n",
" x = list()\n",
" y = list()\n",
" for partitions in MLP_LEVELS:\n",
" tmp = df_graph[(df_graph.Area == area) & (df_graph.Query == query) & (df_graph.MLP_method == mlp) & (df_graph.Levels == \"_\".join(map(str, partitions)))]\n",
" x.append(partitions[0])\n",
" y.append(tmp[\"amount_used_edges\"])\n",
"# speedups.append({\"Query\": query, \"MLP\": mlp, \"_\".join(map(str, partitions)): dijkstra / tmp[\"time\"].mean()})\n",
" plt.plot(x, y, marker=plot_get(query), color=plot_get(mlp), label=query + \"-\" + mlp, alpha=0.7)\n",
"plt.xlabel(\"MLP-Partition-Size\")\n",
"plt.ylabel(\"edges per algorithm\")\n",
"ax.set_yscale('log')\n",
"plt.legend(loc='upper left')\n",
"fig.savefig(\"used-edges-single-level.pgf\", bbox_inches=\"tight\")"
"df_graph.groupby([\"MLP_method\", \"Levels\", \"Query\"])[\"amount_used_edges\"].first()"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "unauthorized-contest",
"id": "humanitarian-armenia",
"metadata": {},
"outputs": [],
"source": [
"df_graph.groupby([\"MLP_method\", \"Levels\", \"Query\"])[\"amount_used_edges\"].first()"
"def format_tex(float_number):\n",
"# exponent = np.floor(np.log10(float_number))\n",
" exponent = 6\n",
" mantissa = float_number/10**exponent\n",
" return \"${:0.1f}\\\\times10^{{{:}}}$\".format(float(mantissa), str(int(exponent)))"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "handy-preservation",
"metadata": {},
"outputs": [],
"source": [
"df_graph[\"amount_used_edges\"] = pd.to_numeric(df_graph[\"amount_used_edges\"], downcast=\"float\")"
]
},
{
Expand All @@ -133,16 +129,15 @@
"for area in AREAS:\n",
" dijkstra = df_graph[(df_graph.Query == \"normal\") & (df_graph.Area == area)][\"amount_used_edges\"].iloc[0]\n",
" print(\"original edge amount:\", dijkstra)\n",
" bidijkstra = df_graph[(df_graph.Query == \"bi\") & (df_graph.Area == area)][\"amount_used_edges\"].iloc[0]\n",
" for query in FAST_QUERY_METHODS:\n",
" for mlp in MLP_METHODS:\n",
" line = dict()\n",
" for partitions in MLP_LEVELS:\n",
" tmp = df_graph[(df_graph.Area == area) & (df_graph.Query == query) & (df_graph.MLP_method == mlp) & (df_graph.Levels == \"_\".join(map(str, partitions)))]\n",
" line[partitions[0]] = tmp[\"amount_used_edges\"].values[0]\n",
" df_new = pd.DataFrame([line])\n",
" df_new[\"MLP_method\"] = mlp\n",
" df_new[\"Query\"] = query\n",
" df_new[\"MLP_method\"] = mlp_title(mlp)\n",
" df_new[\"Query\"] = query.upper()\n",
" df_table = pd.concat([df_table, df_new], ignore_index=True)"
]
},
Expand All @@ -154,7 +149,8 @@
"outputs": [],
"source": [
"df_edges = df_table.groupby([\"Query\", \"MLP_method\"]).first()\n",
"latex = df_edges.to_latex(float_format=\"{:0.1f}\".format)\n",
"# latex = df_edges.to_latex(float_format=\"{:0.1f}\".format)\n",
"latex = df_edges.to_latex(float_format=format_tex, escape=False)\n",
"df_edges"
]
},
Expand All @@ -180,16 +176,162 @@
"latex_list.insert(len(latex_list)-8, '\\midrule')\n",
"latex_list.insert(len(latex_list)-5, '\\midrule')\n",
"latex_new = '\\n'.join(latex_list)\n",
"with open(\"edges.tex\", \"w\") as latex_file:\n",
"latex_new = latex_new.replace(\"MLP_method\", \"MLP-method\")\n",
"with open(OUTPUT + \"-edges.tex\", \"w\") as latex_file:\n",
" latex_file.writelines(latex_new)"
]
},
{
"cell_type": "markdown",
"id": "approximate-belgium",
"metadata": {},
"source": [
"# level"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "funny-tennis",
"metadata": {},
"outputs": [],
"source": [
"OUTPUT = \"level\"\n",
"EVAL_DIR = \"/home/felix/todo/algohol/level\"\n",
"MLP_METHODS = [\"merge\"]\n",
"MLP_LEVELS = [[int(2 ** i)] for i in np.arange(9.0, 11.5, 1.0)]\n",
"MLP_LEVELS = MLP_LEVELS + [[int(2 ** i), 4] for i in np.arange(9.0, 11.5, 1.0)]\n",
"MLP_LEVELS = MLP_LEVELS + [[int(2 ** i), 16] for i in np.arange(9.0, 11.5, 1.0)]\n",
"MLP_LEVELS = MLP_LEVELS + [[int(2 ** i), 4, 4] for i in np.arange(9.0, 11.5, 1.0)]\n",
"FAST_QUERY_METHODS = [\"pcrp\", \"pch\", \"prp\"]\n",
"QUERY_METHODS = [\"normal\"] + FAST_QUERY_METHODS\n",
"AREAS = [\"baden-wuerttemberg\"]\n",
"print(MLP_METHODS, \"with\", MLP_LEVELS)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "bottom-bottom",
"metadata": {},
"outputs": [],
"source": [
"df_graph = pd.DataFrame()\n",
"for area in AREAS:\n",
" for mlp_method in MLP_METHODS:\n",
" for level in MLP_LEVELS:\n",
" for query in QUERY_METHODS:\n",
" df_new = pd.read_json(EVAL_DIR + \"/\" + area + \"-\" + mlp_method + \"-\" + \"_\".join(map(str, level)) + \"-\" + query + \"-info.json\", typ='series')\n",
" df_new = pd.DataFrame([df_new])\n",
" df_new[\"Area\"] = area\n",
" df_new[\"MLP_method\"] = mlp_method\n",
" df_new[\"Levels\"] = \"_\".join(map(str, level))\n",
" df_new[\"Query\"] = query\n",
" df_graph = pd.concat([df_graph, df_new], ignore_index=True)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "champion-spray",
"metadata": {},
"outputs": [],
"source": [
"assert((df_graph.groupby([\"amount_edges\"]).size() == len(QUERY_METHODS)).all())"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "popular-birth",
"metadata": {},
"outputs": [],
"source": [
"df_graph[\"amount_used_edges\"] = pd.to_numeric(df_graph[\"amount_used_edges\"], downcast=\"float\")"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "accessible-courage",
"metadata": {},
"outputs": [],
"source": [
"df_table = pd.DataFrame()\n",
"for area in AREAS:\n",
" dijkstra = df_graph[(df_graph.Query == \"normal\") & (df_graph.Area == area)][\"amount_used_edges\"].iloc[0]\n",
" print(\"original edge amount:\", dijkstra)\n",
" for mlp in MLP_METHODS:\n",
" for partitions in MLP_LEVELS:\n",
" line = dict()\n",
" for query in FAST_QUERY_METHODS:\n",
" tmp = df_graph[(df_graph.Area == area) & (df_graph.Query == query) & (df_graph.MLP_method == mlp) & (df_graph.Levels == \"_\".join(map(str, partitions)))]\n",
" line[query.upper()] = tmp[\"amount_used_edges\"].values[0]\n",
" df_new = pd.DataFrame([line])\n",
" df_new[\"partitions\"] = \"_\".join(map(str, partitions))\n",
" df_table = pd.concat([df_table, df_new], ignore_index=True)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "blind-saint",
"metadata": {},
"outputs": [],
"source": [
"def special_hacky_sort(x):\n",
" splited = x.str.split(\"-\", expand=True)\n",
" return pd.DataFrame(splited).astype(float).sum(axis=1)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "attached-physics",
"metadata": {},
"outputs": [],
"source": [
"df_table['partitions'] = df_table['partitions'].str.replace('_','-')\n",
"df_edges = df_table.groupby(\"partitions\").first()\n",
"df_edges = df_edges.sort_values(by=\"partitions\", key=special_hacky_sort)\n",
"latex = df_edges.to_latex(float_format=format_tex, escape=False)\n",
"df_edges"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "confidential-decade",
"metadata": {},
"outputs": [],
"source": [
"# fixup ugly latex code to have single line header\n",
"latex_list = latex.splitlines()\n",
"\n",
"latex_list[0] = latex_list[0].replace('lr', 'l|r', 1)\n",
"\n",
"columns = latex_list[2].split(\"&\")\n",
"indices = latex_list[3].split(\"&\")\n",
"\n",
"latex_list[2] = \" & \\multicolumn{\" + str(len(FAST_QUERY_METHODS)) + \"}{c}{Dijkstra-Query} \\\\\\\\\"\n",
"\n",
"latex_list[3] = \"&\".join(indices[:1] + columns[1:])\n",
"\n",
"\n",
"latex_list.insert(len(latex_list)-10, '\\midrule')\n",
"latex_list.insert(len(latex_list)-6, '\\midrule')\n",
"latex_new = '\\n'.join(latex_list)\n",
"\n",
"with open(OUTPUT + \"-edges.tex\", \"w\") as latex_file:\n",
" latex_file.writelines(latex_new)"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "fresh-nitrogen",
"metadata": {},
"outputs": [],
"source": []
}
],
Expand All @@ -209,7 +351,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.9"
"version": "3.8.11"
}
},
"nbformat": 4,
Expand Down
53 changes: 53 additions & 0 deletions analysis/helper.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,22 +14,43 @@
markers = ['s', 'X', '*', '+', 'o', '^']

identifiert = dict()
mapping_colors = dict()

for method, color in zip(MLP_METHODS, colors):
identifiert[method] = color

for method, marker in zip(QUERY_METHODS, markers):
identifiert[method] = marker

mapping_colors["pch"] = "kmeans"
mapping_colors["pcrp"] = "gonzalez"
mapping_colors["prp"] = "merge"

TEXT_WIDTH = 426.0


def ns_to_ms(value):
return value / 1e6


def sec_to_min(value):
return value / 60


def plot_get(method):
return identifiert[method]


def plot_color_get(method):
return identifiert[mapping_colors[method]]


def mlp_title(method):
if method == "kmeans":
return "K-means"
return method.title()


def shell_execute(command, EVAL_DIR):
start_time = time.time()
result = subprocess.run(command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, universal_newlines=True)
Expand Down Expand Up @@ -75,6 +96,38 @@ def not_created_yet(file, EVAL_DIR):
return False


def set_pgf_size(width_pt, fraction=1, subplots=(1, 1)):
"""Set figure dimensions to sit nicely in our document.
Parameters
----------
width_pt: float
Document width in points
fraction: float, optional
Fraction of the width which you wish the figure to occupy
subplots: array-like, optional
The number of rows and columns of subplots.
Returns
-------
fig_dim: tuple
Dimensions of figure in inches
"""
# Width of figure (in pts)
fig_width_pt = width_pt * fraction
# Convert from pt to inches
inches_per_pt = 1 / 72.27

# Golden ratio to set aesthetic figure height
golden_ratio = (5**.5 - 1) / 2

# Figure width in inches
fig_width_in = fig_width_pt * inches_per_pt
# Figure height in inches
fig_height_in = fig_width_in * golden_ratio * (subplots[0] / subplots[1])

return (fig_width_in, fig_height_in)


def main():
pass

Expand Down
Loading

0 comments on commit 481a395

Please sign in to comment.