From f249ee66243df94523da882bc5604debb8bfc40a Mon Sep 17 00:00:00 2001
From: Alexander Cowen-Rivers <mc_rivers@icloud.com>
Date: Mon, 17 Jan 2022 18:40:34 +0800
Subject: [PATCH] Update optimisation.ipynb

Update mace-->hebo
---
 HEBO/doc/source/optimisation.ipynb | 40 +++++++++++++++---------------
 1 file changed, 20 insertions(+), 20 deletions(-)

diff --git a/HEBO/doc/source/optimisation.ipynb b/HEBO/doc/source/optimisation.ipynb
index 8c9bc6b9..0aa5c69e 100644
--- a/HEBO/doc/source/optimisation.ipynb
+++ b/HEBO/doc/source/optimisation.ipynb
@@ -140,8 +140,8 @@
     "\n",
     "We compare three BO algorithms:\n",
     "- The basic BO with LCB acquisition\n",
-    "- MACE where three acquisition functions are ensembled, with input warping and power transformation\n",
-    "- MACE with parallel recommendation\n",
+    "- HEBO where three acquisition functions are ensembled, with input warping and power transformation\n",
+    "- HEBO with parallel recommendation\n",
     "\n",
     "We'll run BO with LCB firstly, we can see the interface is very simple"
    ]
@@ -187,7 +187,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "We then run the sequential MACE"
+    "We then run the sequential HEBO"
    ]
   },
   {
@@ -219,12 +219,12 @@
     }
    ],
    "source": [
-    "mace_seq   = HEBO(space, model_name = 'gpy', rand_sample = 4)\n",
+    "hebo_seq   = HEBO(space, model_name = 'gpy', rand_sample = 4)\n",
     "for i in range(64):\n",
-    "    rec_x = mace_seq.suggest(n_suggestions=1)\n",
-    "    mace_seq.observe(rec_x, obj(rec_x))\n",
+    "    rec_x = hebo_seq.suggest(n_suggestions=1)\n",
+    "    hebo_seq.observe(rec_x, obj(rec_x))\n",
     "    if i % 4 == 0:\n",
-    "        print('Iter %d, best_y = %.2f' % (i, mace_seq.y.min()))"
+    "        print('Iter %d, best_y = %.2f' % (i, hebo_seq.y.min()))"
    ]
   },
   {
@@ -263,11 +263,11 @@
     }
    ],
    "source": [
-    "mace_batch = HEBO(space, model_name = 'gpy', rand_sample = 4)\n",
+    "hebo_batch = HEBO(space, model_name = 'gpy', rand_sample = 4)\n",
     "for i in range(16):\n",
-    "    rec_x = mace_batch.suggest(n_suggestions=8)\n",
-    "    mace_batch.observe(rec_x, obj(rec_x))\n",
-    "    print('Iter %d, best_y = %.2f' % (i, mace_batch.y.min()))"
+    "    rec_x = hebo_batch.suggest(n_suggestions=8)\n",
+    "    hebo_batch.observe(rec_x, obj(rec_x))\n",
+    "    print('Iter %d, best_y = %.2f' % (i, hebo_batch.y.min()))"
    ]
   },
   {
@@ -276,8 +276,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "conv_mace_batch = np.minimum.accumulate(mace_batch.y)\n",
-    "conv_mace_seq   = np.minimum.accumulate(mace_seq.y)\n",
+    "conv_hebo_batch = np.minimum.accumulate(hebo_batch.y)\n",
+    "conv_hebo_seq   = np.minimum.accumulate(hebo_seq.y)\n",
     "conv_bo_seq     = np.minimum.accumulate(bo.y)"
    ]
   },
@@ -285,7 +285,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "We plot the converge below, we can see it seems that the LCB converges slightly faster than MACE, while batched MACE used most function evaluations"
+    "We plot the converge below, we can see it seems that the LCB converges slightly faster than HEBO, while batched HEBO used most function evaluations"
    ]
   },
   {
@@ -318,8 +318,8 @@
    ],
    "source": [
     "plt.figure(figsize = (8,6))\n",
-    "plt.semilogy(conv_mace_batch - problem.ideal_point(), 'x-',label = 'MACE, Parallel,Batch = 8')\n",
-    "plt.semilogy(conv_mace_seq   - problem.ideal_point(), 'x-',label = 'MACE, Sequential')\n",
+    "plt.semilogy(conv_hebo_batch - problem.ideal_point(), 'x-',label = 'HEBO, Parallel,Batch = 8')\n",
+    "plt.semilogy(conv_hebo_seq   - problem.ideal_point(), 'x-',label = 'HEBO, Sequential')\n",
     "plt.semilogy(conv_bo_seq     - problem.ideal_point(), 'x-',label = 'BO, LCB')\n",
     "\n",
     "plt.xlabel('Evaluation')\n",
@@ -331,7 +331,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Batched MACE used most function evaluations because it recommend eight points per iterations, we can also plot the converge against **iterations** instead of **evaluations**"
+    "Batched HEBO used most function evaluations because it recommend eight points per iterations, we can also plot the converge against **iterations** instead of **evaluations**"
    ]
   },
   {
@@ -364,8 +364,8 @@
    ],
    "source": [
     "plt.figure(figsize = (8,6))\n",
-    "plt.semilogy(conv_mace_batch[::8] - problem.ideal_point(), 'x-',label = 'MACE, Parallel,Batch = 8')\n",
-    "plt.semilogy(conv_mace_seq   - problem.ideal_point(), 'x-',label = 'MACE, Sequential')\n",
+    "plt.semilogy(conv_hebo_batch[::8] - problem.ideal_point(), 'x-',label = 'HEBO, Parallel,Batch = 8')\n",
+    "plt.semilogy(conv_hebo_seq   - problem.ideal_point(), 'x-',label = 'HEBO, Sequential')\n",
     "plt.semilogy(conv_bo_seq     - problem.ideal_point(), 'x-',label = 'BO, LCB')\n",
     "\n",
     "plt.xlabel('Iterations')\n",
@@ -377,7 +377,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "We can see that parallel MACE converges much faster"
+    "We can see that parallel HEBO converges much faster"
    ]
   }
  ],