"Let's now switch to stochastic gradient descent training instead, which is much faster.\n",
"\n",
"The graph will be similar, except that instead of holding all the training data into a constant node, we create a `Placeholder` node which will be fed actual data at every call of `sesion.run()`."
"The graph will be similar, except that instead of holding all the training data into a constant node, we create a `Placeholder` node which will be fed actual data at every call of `session.run()`."
]
},
{
...
...
@@ -577,7 +577,7 @@
"Problem\n",
"-------\n",
"\n",
"Turn the logistic regression example with SGD into a 1-hidden layer neural network with rectified linear units (nn.relu()) and 1024 hidden nodes. This model should improve your validation / test accuracy.\n",
"Turn the logistic regression example with SGD into a 1-hidden layer neural network with rectified linear units [nn.relu()](https://www.tensorflow.org/versions/r0.7/api_docs/python/nn.html#relu) and 1024 hidden nodes. This model should improve your validation / test accuracy.\n",
" # Initialize all the variables defined above.\n",
...
...
@@ -617,7 +617,7 @@
" # We are going to minimize the L2 loss. The L2 loss is the sum of the\n",
" # squared error for all our estimates of y. This penalizes large errors\n",
" # a lot, but small errors only a little.\n",
" loss = tf.reduce_mean(tf.nn.l2_loss(yerror))\n",
" loss = tf.nn.l2_loss(yerror)\n",
"\n",
" # Perform gradient descent. \n",
" # This essentially just updates weights, like weights += grads * mu\n",
...
...
@@ -824,9 +824,9 @@
"\n",
"The first line calculates the L2 loss manually. It's the same as `l2_loss(yerror)`, which is half of the sum of the squared error, so $\\frac{1}{2} \\sum (\\hat{y} - y)^2$. With this code, you can see exactly what the `l2_loss` operation does. It's the total of all the squared differences between the target and our estimates. And minimizing the L2 loss will minimize how much our estimates of $y$ differ from the true values of $y$.\n",
"\n",
"The second line calculates $\\sum{x_i (\\hat{y} - y)}$. What is that? It's the partial derivative of the L2 loss, the same thing as what `gradients(loss, weights)` does in the earlier code. Not sure about that? Let's look at it in more detail. The gradient calculation is going to get the partial derivatives of loss with respect to each of the weights so we can change those weights in the direction that will reduce the loss. L2 loss is $\\frac{1}{2} \\sum (\\hat{y} - y)^2$, where $\\hat{y} = w_2 x + w_1$. So, using the chain rule and substituting in for $\\hat{y}$ in the derivative, $\\frac{\\partial}{\\partial w_i} = \\sum{(\\hat{y} - y)\\, x_i}$. `GradientDescentOptimizer` does these calculations automatically for you based on the graph structure.\n",
"The second line calculates $\\begin{bmatrix}\\sum{(\\hat{y} - y)*1} \\\\ \\sum{(\\hat{y} - y)*x_i}\\end{bmatrix}$. What is that? It's the partial derivatives of the L2 loss with respect to $w_1$ and $w_2$, the same thing as what `gradients(loss, weights)` does in the earlier code. Not sure about that? Let's look at it in more detail. The gradient calculation is going to get the partial derivatives of loss with respect to each of the weights so we can change those weights in the direction that will reduce the loss. L2 loss is $\\frac{1}{2} \\sum (\\hat{y} - y)^2$, where $\\hat{y} = w_2 x + w_1$. So, using the chain rule and substituting in for $\\hat{y}$ in the derivative, $\\frac{\\partial}{\\partial w_2} = \\sum{(\\hat{y} - y)\\, *x_i}$ and $\\frac{\\partial}{\\partial w_1} = \\sum{(\\hat{y} - y)\\, *1}$. `GradientDescentOptimizer` does these calculations automatically for you based on the graph structure.\n",
"\n",
"The third line is equivalent to `weights -= mu * gradient`, so it subtracts a constant the gradient after scaling by the learning rate (to avoid jumping too far each time, which risks moving in the wrong direction). It's also the same thing that `GradientDescentOptimizer(learning_rate).minimize(loss)` does in the earlier code. Gradient descient updates its first parameter based on the values in the second after scaling by the third, so it's equivalent to the `assign_sub(weights, mu * gradient)`.\n",
"The third line is equivalent to `weights -= mu * gradient`, so it subtracts a constant the gradient after scaling by the learning rate (to avoid jumping too far each time, which risks moving in the wrong direction). It's also the same thing that `GradientDescentOptimizer(learning_rate).minimize(loss)` does in the earlier code. Gradient descent updates its first parameter based on the values in the second after scaling by the third, so it's equivalent to the `assign_sub(weights, mu * gradient)`.\n",
"\n",
"Hopefully, this other code gives you a better understanding of what the operations we used previously are actually doing. In practice, you'll want to use those high level operators most of the time rather than calculating things yourself. For this toy example and simple network, it's not too bad to compute and apply the gradients yourself from scratch, but things get more complicated with larger networks."
