Merge pull request #93 from KellenSunderland/P05_C01

Fixed some typos and a bug in P05-C01
zackchase · Aug 7, 2017 · 0536450 · 0536450
2 parents 1dcd2cf + 92ad07c
commit 0536450
Showing 1 changed file with 15 additions and 15 deletions.
diff --git a/P05-C01-simple-rnn.ipynb b/P05-C01-simple-rnn.ipynb
@@ -58,7 +58,7 @@
    "source": [
     "## Dataset: \"The Time Machine\" \n",
     "\n",
-    "Now mess with some data. I grabbed a copy of the ``Time Machine``, mostly because it's available freely thanks to the good people at [Project Gutenberg](http://www.gutenberg.org) and a lot of people are tired of seeing RNNs generate Shakespeare. In case you prefer to torturing Shakespeare to torturing H.G. Wells, I've also included Andrej Karpathy's tinyshakespeare.txt in the data folder. Let's get started by reading in the data."
+    "Now mess with some data. I grabbed a copy of the ``Time Machine``, mostly because it's available freely thanks to the good people at [Project Gutenberg](http://www.gutenberg.org) and a lot of people are tired of seeing RNNs generate Shakespeare. In case you prefer torturing Shakespeare to torturing H.G. Wells, I've also included Andrej Karpathy's tinyshakespeare.txt in the data folder. Let's get started by reading in the data."
    ]
   },
   {
@@ -263,7 +263,7 @@
    "source": [
     "## One-hot representations\n",
     "\n",
-    "We can use NDArray's one_hot() render a one-hot representation of each character. But frack it, since this is the from scratch tutorial, let's right this ourselves."
+    "We can use NDArray's one_hot() operation to render a one-hot representation of each character. But frack it, since this is the from scratch tutorial, let's write this ourselves."
    ]
   },
   {
@@ -431,7 +431,7 @@
     "sequences_per_batch_row = int(np.floor(len(dataset))/batch_size)\n",
     "print(sequences_per_batch_row)\n",
     "data_rows = [dataset[i*sequences_per_batch_row:i*sequences_per_batch_row+sequences_per_batch_row] \n",
-    "            for i in range(batch_size)]"
+    "             for i in range(batch_size)]"
    ]
   },
   {
@@ -440,7 +440,7 @@
     "collapsed": true
    },
    "source": [
-    "Let's sanity check that everything went the way we hop. For each data_row, the second sequence should follow the first:"
+    "Let's sanity check that everything went the way we hope. For each data_row, the second sequence should follow the first:"
    ]
   },
   {
@@ -497,9 +497,9 @@
     "    # iterate over the sequences\n",
     "    for s in range(len(datasets[0])):\n",
     "        sequence = []\n",
-    "        # iterate over the elements of the seqeunce\n",
+    "        # iterate over the elements of the sequence\n",
     "        for elem in range(len(datasets[0][0])):\n",
-    "            sequence.append(nd.concatenate([ds[s][elem].reshape((1,-1)) for ds in datasets], axis=0))\n",
+    "            sequence.append(nd.concatenate([ds[s][elem].reshape((1, -1)) for ds in datasets], axis=0))\n",
     "        full_dataset.append(sequence)\n",
     "    return(full_dataset)\n",
     "        "
@@ -542,8 +542,8 @@
    ],
    "source": [
     "print(training_data[0][0].shape)\n",
-    "print(\"Seq 0, Batch 0 \\\"%s\\\"\" % textify([training_data[0][i][0].reshape((1,-1)) for i in range(seq_length)]))\n",
-    "print(\"Seq 1, Batch 0 \\\"%s\\\"\" % textify([training_data[1][i][0].reshape((1,-1)) for i in range(seq_length)]))"
+    "print(\"Seq 0, Batch 0 \\\"%s\\\"\" % textify([training_data[0][i][0].reshape((1, -1)) for i in range(seq_length)]))\n",
+    "print(\"Seq 1, Batch 0 \\\"%s\\\"\" % textify([training_data[1][i][0].reshape((1, -1)) for i in range(seq_length)]))"
    ]
   },
   {
@@ -601,8 +601,8 @@
     }
    ],
    "source": [
-    "print(textify([training_data[0][i][2].reshape((1,-1)) for i in range(seq_length)]))\n",
-    "print(textify([training_labels[0][i][2].reshape((1,-1)) for i in range(seq_length)]))"
+    "print(textify([training_data[0][i][2].reshape((1, -1)) for i in range(seq_length)]))\n",
+    "print(textify([training_labels[0][i][2].reshape((1, -1)) for i in range(seq_length)]))"
    ]
   },
   {
@@ -737,9 +737,9 @@
    ],
    "source": [
     "####################\n",
-    "# With a temperature of 1 (always 1 during training), we get back some set of proabilities\n",
+    "# With a temperature of 1 (always 1 during training), we get back some set of probabilities\n",
     "####################\n",
-    "softmax(nd.array([[1,-1],[-1,1]]), temperature=1000.0)"
+    "softmax(nd.array([[1, -1], [-1, 1]]), temperature=1.0)"
    ]
   },
   {
@@ -751,8 +751,8 @@
      "data": {
       "text/plain": [
        "\n",
-       "[[ 0.50049996  0.49949998]\n",
-       " [ 0.49949998  0.50049996]]\n",
+       "[[ 0.88079703  0.11920292]\n",
+       " [ 0.11920292  0.88079703]]\n",
        "<NDArray 2x2 @cpu(0)>"
       ]
      },
@@ -789,7 +789,7 @@
    ],
    "source": [
     "####################\n",
-    "# Often we want to sample with low temperatures to produce sharp proababilities\n",
+    "# Often we want to sample with low temperatures to produce sharp probabilities\n",
     "####################\n",
     "softmax(nd.array([[10,-10],[-10,10]]), temperature=.1)"
    ]