From 19c8df6b720e80b77cf769fa98703bc2a341f7b0 Mon Sep 17 00:00:00 2001
From: Bas Nijholt <basnijholt@gmail.com>
Date: Wed, 25 Sep 2019 14:12:57 +0200
Subject: [PATCH] work on figures
---
figures.ipynb | 35 +++++++++++++++++++----------------
phase_diagram.ipynb | 7 ++++---
2 files changed, 23 insertions(+), 19 deletions(-)
diff --git a/figures.ipynb b/figures.ipynb
index 3fa61dd..0a19f1d 100644
--- a/figures.ipynb
+++ b/figures.ipynb
@@ -547,7 +547,7 @@
"from matplotlib.patches import Polygon\n",
"\n",
"fig, axs = plt.subplots(5, 1, figsize=(fig_width, 1.5 * fig_height))\n",
- "f = lambda x: np.sin(x) ** 2\n",
+ "f = lambda x: np.sin(np.array(x)) ** 2\n",
"xs = np.array([0, 1.3, 3, 5, 7, 8])\n",
"ys = f(xs)\n",
"\n",
@@ -849,20 +849,22 @@
")\n",
"\n",
"\n",
- "def isoline_loss_function(y_iso, sigma, priority=1):\n",
- " from adaptive.learner.learnerND import default_loss\n",
- "\n",
- " def gaussian(x, mu, sigma):\n",
- " return np.exp(-(x - mu) ** 2 / sigma ** 2)\n",
+ "from adaptive.learner.learnerND import default_loss\n",
"\n",
- " def loss(simplex, ys):\n",
- " distance = np.mean([abs(y_iso - y) for y in ys])\n",
- " return priority * gaussian(distance, 0, sigma) + default_loss(simplex, ys)\n",
+ "def gaussian(x, mu, sigma):\n",
+ " return np.exp(-(x - mu)**2 / sigma**2 / 2)\n",
"\n",
+ "def isoline_loss_function(level, sigma, priority):\n",
+ " def loss(simplex, values, value_scale):\n",
+ " values = np.array(values)\n",
+ " dist = abs(level * value_scale - values).mean()\n",
+ " L_default = default_loss(simplex, values, value_scale)\n",
+ " L_dist = priority * gaussian(dist, 0, sigma)\n",
+ " return L_dist + L_default\n",
" return loss\n",
"\n",
- "\n",
- "loss = isoline_loss_function(y_iso=0.1, sigma=1, priority=0.0)\n",
+ "level = 0.1\n",
+ "loss_per_simplex = isoline_loss_function(level, 0.4, 0.5)\n",
"\n",
"learners = []\n",
"fnames = []\n",
@@ -907,17 +909,18 @@
"npoints = 17 ** 2\n",
"\n",
"\n",
- "def isoline_loss_function(y_iso, sigma, priority=1):\n",
+ "def isoline_loss_function(level, sigma, priority=1):\n",
" from adaptive.learner.learnerND import default_loss\n",
"\n",
" def gaussian(x, mu, sigma):\n",
" return np.exp(-(x - mu) ** 2 / sigma ** 2 / 2)\n",
"\n",
" def loss(simplex, values, value_scale):\n",
- " distance = np.mean([abs(y_iso * value_scale - y) for y in values])\n",
- " return priority * gaussian(distance, 0, sigma) + default_loss(\n",
- " simplex, values, value_scale\n",
- " )\n",
+ " values = np.array(values)\n",
+ " dist = abs(level * value_scale - values).mean()\n",
+ " L_default = default_loss(simplex, values, value_scale)\n",
+ " L_dist = priority * gaussian(dist, 0, sigma)\n",
+ " return L_dist + L_default\n",
"\n",
" return loss\n",
"\n",
diff --git a/phase_diagram.ipynb b/phase_diagram.ipynb
index 1f0e82f..b304978 100644
--- a/phase_diagram.ipynb
+++ b/phase_diagram.ipynb
@@ -115,8 +115,9 @@
" return learner.npoints > 100_000\n",
"\n",
"scheduler = adaptive_scheduler.scheduler.PBS(\n",
- " cores=13*10,\n",
- " cores_per_node=10,\n",
+ " cores=10*20,\n",
+ " cores_per_node=20,\n",
+ " executor_type='ipyparallel',\n",
") # every learner get this many cores\n",
"\n",
"run_manager = adaptive_scheduler.server_support.RunManager(\n",
@@ -124,7 +125,7 @@
" learners_file=\"learners_file.py\",\n",
" goal=goal,\n",
" log_interval=30,\n",
- " save_interval=300,\n",
+ " save_interval=120,\n",
")\n",
"run_manager.start()\n"
]
--
GitLab