Skip to content

Commit

Permalink
Deploying to gh-pages from @ ca445e2 🚀
Browse files Browse the repository at this point in the history
  • Loading branch information
@danhalligan
danhalligan committed Aug 22, 2024
1 parent a9c1456 commit dac8710
Show file tree
Hide file tree
Showing 12 changed files with 79 additions and 75 deletions.
4 changes: 2 additions & 2 deletions 03-linear-regression.md
View file

Large diffs are not rendered by default.

2 changes: 1 addition & 1 deletion 05-resampling-methods.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -170,7 +170,7 @@ mean(store)
```
```
## [1] 0.6307
## [1] 0.6294
```

The probability of including $4$ when resampling numbers $1...100$ is close to
Expand Down
2 changes: 1 addition & 1 deletion 08-tree-based-methods.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -509,7 +509,7 @@ bartfit <- gbart(Carseats[train, 2:11], Carseats[train, 1],
## done 800 (out of 1100)
## done 900 (out of 1100)
## done 1000 (out of 1100)
## time: 3s
## time: 2s
## trcnt,tecnt: 1000,1000
```

Expand Down
62 changes: 31 additions & 31 deletions 10-deep-learning.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -393,15 +393,15 @@ npred <- predict(nn, x[testid, ])
```

```
## 6/6 - 0s - 54ms/epoch - 9ms/step
## 6/6 - 0s - 53ms/epoch - 9ms/step
```

``` r
mean(abs(y[testid] - npred))
```

```
## [1] 2.30611
## [1] 2.324471
```

In this case, the neural network outperforms logistic regression having a lower
Expand Down Expand Up @@ -433,7 +433,7 @@ model <- application_resnet50(weights = "imagenet")

```
## Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50_weights_tf_dim_ordering_tf_kernels.h5
## 8192/102967424 [..............................] - ETA: 0s 4235264/102967424 [>.............................] - ETA: 1s 18759680/102967424 [====>.........................] - ETA: 0s 36265984/102967424 [=========>....................] - ETA: 0s 53231616/102967424 [==============>...............] - ETA: 0s 70893568/102967424 [===================>..........] - ETA: 0s 86900736/102967424 [========================>.....] - ETA: 0s102967424/102967424 [==============================] - 0s 0us/step
## 8192/102967424 [..............................] - ETA: 0s 2531328/102967424 [..............................] - ETA: 1s 15474688/102967424 [===>..........................] - ETA: 0s 26615808/102967424 [======>.......................] - ETA: 0s 41254912/102967424 [===========>..................] - ETA: 0s 56344576/102967424 [===============>..............] - ETA: 0s 70606848/102967424 [===================>..........] - ETA: 0s 85114880/102967424 [=======================>......] - ETA: 0s 99860480/102967424 [============================>.] - ETA: 0s102967424/102967424 [==============================] - 0s 0us/step
```

``` r
Expand Down Expand Up @@ -729,7 +729,7 @@ kpred <- predict(model, xrnn[!istrain,, ])
```

```
## [1] 0.4118947
## [1] 0.4126973
```

Both models estimate the same number of coefficients/weights (16):
Expand Down Expand Up @@ -762,25 +762,25 @@ model$get_weights()

```
## [[1]]
## [,1]
## [1,] -0.030813020
## [2,] 0.101270206
## [3,] 0.179615468
## [4,] -0.006129819
## [5,] 0.124265596
## [6,] -0.068290100
## [7,] 0.037529659
## [8,] 0.077909760
## [9,] 0.204727829
## [10,] -0.032089159
## [11,] 0.034063924
## [12,] -0.847255647
## [13,] 0.096230716
## [14,] 0.511620998
## [15,] 0.529515982
## [,1]
## [1,] -0.03103472
## [2,] 0.09990595
## [3,] 0.16059186
## [4,] -0.00529294
## [5,] 0.12038019
## [6,] -0.04571687
## [7,] 0.03948567
## [8,] 0.07816564
## [9,] 0.18490496
## [10,] -0.02650242
## [11,] 0.03655238
## [12,] -0.83607799
## [13,] 0.10024448
## [14,] 0.51236451
## [15,] 0.52258915
##
## [[2]]
## [1] -0.006476609
## [1] -0.004416319
```

The flattened RNN has a lower $R^2$ on the test data than our `lm` model
Expand Down Expand Up @@ -833,11 +833,11 @@ xfun::cache_rds({
```

```
## 56/56 - 0s - 63ms/epoch - 1ms/step
## 56/56 - 0s - 64ms/epoch - 1ms/step
```

```
## [1] 0.4221822
## [1] 0.4276608
```

This approach improves our $R^2$ over the linear model above.
Expand Down Expand Up @@ -910,7 +910,7 @@ xfun::cache_rds({
```

```
## [1] 0.451564
## [1] 0.4472675
```

### Question 13
Expand Down Expand Up @@ -966,9 +966,9 @@ xfun::cache_rds({

```
## Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/imdb.npz
## 8192/17464789 [..............................] - ETA: 0s 4202496/17464789 [======>.......................] - ETA: 0s17464789/17464789 [==============================] - 0s 0us/step
## 782/782 - 15s - 15s/epoch - 20ms/step
## 782/782 - 15s - 15s/epoch - 20ms/step
## 8192/17464789 [..............................] - ETA: 0s 3227648/17464789 [====>.........................] - ETA: 0s 4202496/17464789 [======>.......................] - ETA: 0s 8396800/17464789 [=============>................] - ETA: 0s17464789/17464789 [==============================] - 0s 0us/step
## 782/782 - 15s - 15s/epoch - 19ms/step
## 782/782 - 15s - 15s/epoch - 19ms/step
## 782/782 - 15s - 15s/epoch - 20ms/step
## 782/782 - 15s - 15s/epoch - 20ms/step
```
Expand All @@ -977,10 +977,10 @@ xfun::cache_rds({

| Max Features| Accuracy|
|------------:|--------:|
| 1000| 0.84476|
| 3000| 0.86616|
| 5000| 0.86156|
| 10000| 0.87264|
| 1000| 0.80568|
| 3000| 0.87264|
| 5000| 0.87600|
| 10000| 0.87420|

Varying the dictionary size does not make a substantial impact on our estimates
of accuracy. However, the models do take a substantial amount of time to fit and
Expand Down
Binary file modified 10-deep-learning_files/figure-html/unnamed-chunk-12-1.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified 10-deep-learning_files/figure-html/unnamed-chunk-21-1.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified 11-survival-analysis-and-censored-data_files/figure-html/unnamed-chunk-7-1.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
74 changes: 39 additions & 35 deletions deep-learning.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -750,9 +750,9 @@ <h3><span class="header-section-number">10.2.2</span> Question 7<a href="deep-le
<span id="cb707-19"><a href="deep-learning.html#cb707-19" aria-hidden="true"></a><span class="kw">plot</span>(history, <span class="dt">smooth =</span> <span class="ot">FALSE</span>)</span></code></pre></div>
<p><img src="10-deep-learning_files/figure-html/unnamed-chunk-12-1.png" width="672" /></p>
<div class="sourceCode" id="cb708"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb708-1"><a href="deep-learning.html#cb708-1" aria-hidden="true"></a>npred &lt;-<span class="st"> </span><span class="kw">predict</span>(nn, x[testid, ])</span></code></pre></div>
<pre><code>## 6/6 - 0s - 54ms/epoch - 9ms/step</code></pre>
<pre><code>## 6/6 - 0s - 53ms/epoch - 9ms/step</code></pre>
<div class="sourceCode" id="cb710"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb710-1"><a href="deep-learning.html#cb710-1" aria-hidden="true"></a><span class="kw">mean</span>(<span class="kw">abs</span>(y[testid] <span class="op">-</span><span class="st"> </span>npred))</span></code></pre></div>
<pre><code>## [1] 2.30611</code></pre>
<pre><code>## [1] 2.324471</code></pre>
<p>In this case, the neural network outperforms logistic regression having a lower
absolute error rate on the test data.</p>
</div>
Expand All @@ -779,12 +779,14 @@ <h3><span class="header-section-number">10.2.3</span> Question 8<a href="deep-le
<pre><code>## Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50_weights_tf_dim_ordering_tf_kernels.h5
##
8192/102967424 [..............................] - ETA: 0s
4235264/102967424 [&gt;.............................] - ETA: 1s
18759680/102967424 [====&gt;.........................] - ETA: 0s
36265984/102967424 [=========&gt;....................] - ETA: 0s
53231616/102967424 [==============&gt;...............] - ETA: 0s
70893568/102967424 [===================&gt;..........] - ETA: 0s
86900736/102967424 [========================&gt;.....] - ETA: 0s
2531328/102967424 [..............................] - ETA: 1s
15474688/102967424 [===&gt;..........................] - ETA: 0s
26615808/102967424 [======&gt;.......................] - ETA: 0s
41254912/102967424 [===========&gt;..................] - ETA: 0s
56344576/102967424 [===============&gt;..............] - ETA: 0s
70606848/102967424 [===================&gt;..........] - ETA: 0s
85114880/102967424 [=======================&gt;......] - ETA: 0s
99860480/102967424 [============================&gt;.] - ETA: 0s
102967424/102967424 [==============================] - 0s 0us/step</code></pre>
<div class="sourceCode" id="cb714"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb714-1"><a href="deep-learning.html#cb714-1" aria-hidden="true"></a>pred &lt;-<span class="st"> </span>model <span class="op">|</span><span class="er">&gt;</span></span>
<span id="cb714-2"><a href="deep-learning.html#cb714-2" aria-hidden="true"></a><span class="st"> </span><span class="kw">predict</span>(x) <span class="op">|</span><span class="er">&gt;</span></span>
Expand Down Expand Up @@ -1003,7 +1005,7 @@ <h3><span class="header-section-number">10.2.5</span> Question 10<a href="deep-l
<div class="sourceCode" id="cb733"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb733-1"><a href="deep-learning.html#cb733-1" aria-hidden="true"></a>kpred &lt;-<span class="st"> </span><span class="kw">predict</span>(model, xrnn[<span class="op">!</span>istrain,, ])</span></code></pre></div>
<pre><code>## 56/56 - 0s - 58ms/epoch - 1ms/step</code></pre>
<div class="sourceCode" id="cb735"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb735-1"><a href="deep-learning.html#cb735-1" aria-hidden="true"></a><span class="dv">1</span> <span class="op">-</span><span class="st"> </span><span class="kw">mean</span>((kpred <span class="op">-</span><span class="st"> </span>arframe[<span class="op">!</span>istrain, <span class="st">&quot;log_volume&quot;</span>])<span class="op">^</span><span class="dv">2</span>) <span class="op">/</span><span class="st"> </span>V0</span></code></pre></div>
<pre><code>## [1] 0.4118947</code></pre>
<pre><code>## [1] 0.4126973</code></pre>
<p>Both models estimate the same number of coefficients/weights (16):</p>
<div class="sourceCode" id="cb737"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb737-1"><a href="deep-learning.html#cb737-1" aria-hidden="true"></a><span class="kw">coef</span>(arfit)</span></code></pre></div>
<pre><code>## (Intercept) L1.DJ_return L1.log_volume L1.log_volatility
Expand All @@ -1022,25 +1024,25 @@ <h3><span class="header-section-number">10.2.5</span> Question 10<a href="deep-l
## -0.017206826 -0.037298183 0.008361380</code></pre>
<div class="sourceCode" id="cb739"><pre class="sourceCode r"><code class="sourceCode r"><span id="cb739-1"><a href="deep-learning.html#cb739-1" aria-hidden="true"></a>model<span class="op">$</span><span class="kw">get_weights</span>()</span></code></pre></div>
<pre><code>## [[1]]
## [,1]
## [1,] -0.030813020
## [2,] 0.101270206
## [3,] 0.179615468
## [4,] -0.006129819
## [5,] 0.124265596
## [6,] -0.068290100
## [7,] 0.037529659
## [8,] 0.077909760
## [9,] 0.204727829
## [10,] -0.032089159
## [11,] 0.034063924
## [12,] -0.847255647
## [13,] 0.096230716
## [14,] 0.511620998
## [15,] 0.529515982
## [,1]
## [1,] -0.03103472
## [2,] 0.09990595
## [3,] 0.16059186
## [4,] -0.00529294
## [5,] 0.12038019
## [6,] -0.04571687
## [7,] 0.03948567
## [8,] 0.07816564
## [9,] 0.18490496
## [10,] -0.02650242
## [11,] 0.03655238
## [12,] -0.83607799
## [13,] 0.10024448
## [14,] 0.51236451
## [15,] 0.52258915
##
## [[2]]
## [1] -0.006476609</code></pre>
## [1] -0.004416319</code></pre>
<p>The flattened RNN has a lower <span class="math inline">\(R^2\)</span> on the test data than our <code>lm</code> model
above. The <code>lm</code> model is quicker to fit and conceptually simpler also
giving us the ability to inspect the coefficients for different variables.</p>
Expand Down Expand Up @@ -1086,8 +1088,8 @@ <h3><span class="header-section-number">10.2.6</span> Question 11<a href="deep-l
<span id="cb741-27"><a href="deep-learning.html#cb741-27" aria-hidden="true"></a> <span class="dv">1</span> <span class="op">-</span><span class="st"> </span><span class="kw">mean</span>((kpred <span class="op">-</span><span class="st"> </span>arframe[<span class="op">!</span>istrain, <span class="st">&quot;log_volume&quot;</span>])<span class="op">^</span><span class="dv">2</span>) <span class="op">/</span><span class="st"> </span>V0</span>
<span id="cb741-28"><a href="deep-learning.html#cb741-28" aria-hidden="true"></a></span>
<span id="cb741-29"><a href="deep-learning.html#cb741-29" aria-hidden="true"></a>})</span></code></pre></div>
<pre><code>## 56/56 - 0s - 63ms/epoch - 1ms/step</code></pre>
<pre><code>## [1] 0.4221822</code></pre>
<pre><code>## 56/56 - 0s - 64ms/epoch - 1ms/step</code></pre>
<pre><code>## [1] 0.4276608</code></pre>
<p>This approach improves our <span class="math inline">\(R^2\)</span> over the linear model above.</p>
</div>
<div id="question-12-4" class="section level3 hasAnchor" number="10.2.7">
Expand Down Expand Up @@ -1150,7 +1152,7 @@ <h3><span class="header-section-number">10.2.7</span> Question 12<a href="deep-l
<span id="cb744-48"><a href="deep-learning.html#cb744-48" aria-hidden="true"></a></span>
<span id="cb744-49"><a href="deep-learning.html#cb744-49" aria-hidden="true"></a>})</span></code></pre></div>
<pre><code>## 56/56 - 0s - 135ms/epoch - 2ms/step</code></pre>
<pre><code>## [1] 0.451564</code></pre>
<pre><code>## [1] 0.4472675</code></pre>
</div>
<div id="question-13-3" class="section level3 hasAnchor" number="10.2.8">
<h3><span class="header-section-number">10.2.8</span> Question 13<a href="deep-learning.html#question-13-3" class="anchor-section" aria-label="Anchor link to header"></a></h3>
Expand Down Expand Up @@ -1203,10 +1205,12 @@ <h3><span class="header-section-number">10.2.8</span> Question 13<a href="deep-l
<pre><code>## Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/imdb.npz
##
8192/17464789 [..............................] - ETA: 0s
3227648/17464789 [====&gt;.........................] - ETA: 0s
4202496/17464789 [======&gt;.......................] - ETA: 0s
8396800/17464789 [=============&gt;................] - ETA: 0s
17464789/17464789 [==============================] - 0s 0us/step
## 782/782 - 15s - 15s/epoch - 20ms/step
## 782/782 - 15s - 15s/epoch - 20ms/step
## 782/782 - 15s - 15s/epoch - 19ms/step
## 782/782 - 15s - 15s/epoch - 19ms/step
## 782/782 - 15s - 15s/epoch - 20ms/step
## 782/782 - 15s - 15s/epoch - 20ms/step</code></pre>
<table>
Expand All @@ -1219,19 +1223,19 @@ <h3><span class="header-section-number">10.2.8</span> Question 13<a href="deep-l
<tbody>
<tr class="odd">
<td align="right">1000</td>
<td align="right">0.84476</td>
<td align="right">0.80568</td>
</tr>
<tr class="even">
<td align="right">3000</td>
<td align="right">0.86616</td>
<td align="right">0.87264</td>
</tr>
<tr class="odd">
<td align="right">5000</td>
<td align="right">0.86156</td>
<td align="right">0.87600</td>
</tr>
<tr class="even">
<td align="right">10000</td>
<td align="right">0.87264</td>
<td align="right">0.87420</td>
</tr>
</tbody>
</table>
Expand Down
Loading

0 comments on commit dac8710

Please sign in to comment.