MachineLearning/hands-on
1
0
Fork 0
Code Issues Pull Requests Releases Activity

made initial weights higher-variance

Browse Source
This commit is contained in:
arodiss
2018-02-10 19:24:56 +02:00
parent 78f7872c6f
commit 3c0e93b3d5
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
extra_capsnets.ipynb
View File

@@ -607,7 +607,7 @@
"cell_type": "markdown", "cell_type": "markdown",
"metadata": {}, "metadata": {},
"source": [ "source": [
"Okay, let's start by creating a trainable variable of shape (1, 1152, 10, 16, 8) that will hold all the transformation matrices. The first dimension of size 1 will make this array easy to tile. We initialize this variable randomly using a normal distribution with a standard deviation to 0.01." "Okay, let's start by creating a trainable variable of shape (1, 1152, 10, 16, 8) that will hold all the transformation matrices. The first dimension of size 1 will make this array easy to tile. We initialize this variable randomly using a normal distribution with a standard deviation to 0.1."
] ]
}, },
{ {
@@ -616,7 +616,7 @@
"metadata": {}, "metadata": {},
"outputs": [], "outputs": [],
"source": [ "source": [
"init_sigma = 0.01\n", "init_sigma = 0.1\n",
"\n", "\n",
"W_init = tf.random_normal(\n", "W_init = tf.random_normal(\n",
" shape=(1, caps1_n_caps, caps2_n_caps, caps2_n_dims, caps1_n_dims),\n", " shape=(1, caps1_n_caps, caps2_n_caps, caps2_n_dims, caps1_n_dims),\n",