I am getting started with deep learning and have a basic question on CNN's.
I understand how gradients are adjusted using backpropagation according to a loss function.
But I thought the values of the convolving filter matrices (in CNN's) needs to be determined by us.
I'm using Keras and this is how (from a tutorial) the convolution layer was defined:
classifier = Sequential()
classifier.add(Conv2D(32, (3, 3), input_shape = (64, 64, 3), activation = 'relu'))
There are 32 filter matrices with dimensions 3x3 is used.
But, how are the values for these 32x3x3 matrices are determined?
It's not the gradients that are adjusted, the gradient calculated with the backpropagation algorithm is just the group of partial derivatives with respect to each weight in the network, and these components are in turn used to adjust the network weights in order to minimize the loss.
Take a look at this introductive guide.
The weights in the convolution layer in your example will be initialized to random values (according to a specific method), and then tweaked during training, using the gradient at each iteration to adjust each individual weight. Same goes for weights in a fully connected layer, or any other layer with weights.
EDIT: I'm adding some more details about the answer above.
Let's say you have a neural network with a single layer, which has some weights W. Now, during the forward pass, you calculate your output yHat for your network, compare it with your expected output y for your training samples, and compute some cost C (for example, using the quadratic cost function).
Now, you're interested in making the network more accurate, ie. you'd like to minimize C as much as possible. Imagine you want to find the minimum value for simple function like f(x)=x^2. You can start at some random point (as you did with your network), then compute the slope of the function at that point (ie, the derivative) and move down that direction, until you reach a minimum value (a local minimum at least).
With a neural network it's the same idea, with the difference that your inputs are fixed (the training samples), and you can see your cost function C as having n variables, where n is the number of weights in your network. To minimize C, you need the slope of the cost function C in each direction (ie. with respect to each variable, each weight w), and that vector of partial derivatives is the gradient.
Once you have the gradient, the part where you "move a bit following the slope" is the weights update part, where you update each network weight according to its partial derivative (in general, you subtract some learning rate multiplied by the partial derivative with respect to that weight).
A trained network is just a network whose weights have been adjusted over many iterations in such a way that the value of the cost function C over the training dataset is as small as possible.
This is the same for a convolutional layer too: you first initialize the weights at random (ie. you place yourself on a random position on the plot for the cost function C), then compute the gradients, then "move downhill", ie. you adjust each weight following the gradient in order to minimize C.
The only difference between a fully connected layer and a convolutional layer is how they calculate their outputs, and how the gradient is in turn computed, but the part where you update each weight with the gradient is the same for every weight in the network.
So, to answer your question, those filters in the convolutional kernels are initially random and are later adjusted with the backpropagation algorithm, as described above.
Hope this helps!
Sergio0694 states ,"The weights in the convolution layer in your example will be initialized to random values". So if they are random and say I want 10 filters. Every execution algorithm could find different filter. Also say I have Mnist data set. Numbers are formed of edges and curves. Is it guaranteed that there will be a edge filter or curve filter in 10?
I mean is first 10 filters most meaningful most distinctive filters we can find.
best
Related
I am using pytorch for evaluating gradients of feed-forward network, but only for a subset of parameters, related to the first two layers.
Since backpropagation is carried backwards layer by layer, I wonder: why is it computationally faster than evaluating gradients of whole network?
Pytorch builds a computation graph for backward propagation that only contains the minimum nodes and edges to get the accumulated gradient for leaves that require gradient. Even if the first two layers require gradient, there are many tensors (intermediate tensors or frozen parameters tensors) that are unused and that are cut in the backward graph. Plus the built-in function AccumulatedGradient that stores the gradients in .grad attribute is call less time reducing the total computation time too.
Here you can see an example for an "AddBackward Node" where for instance A is an intermediate tensor computed with the first two layers and B is the 3rd (constant) layer that can be ignored.
An other example: if you have a matrix-matrix product (MmBackward Node) that uses an intermediate tensor that not depends on the 2 first layers. In this case the tensor itself is required to compute the backprop but the "previous" tensors that were used to compute it can be ignored in the graph.
To visualize the sub-graph that is actually computed (and compare when the model is unfrozen), you can use torchviz.
i am currently reading " Network in Network' paper.
And in the paper, it is stated that
"the cross channel parametric pooling layer is also equivalent to convolution layer with
1x1 convolution kernel. "
My question is first of all, what is cross channel parametric pooling layer exactly mean?is it just fully connected layer?
And why is cross channel parametric pooling layer same with 1x1 convolution kernel.
It would be thankful if you answer both mathematically and with examples.
Please help me~
I haven't read the paper but I have a fair idea of what this is. First of all
How is a 1x1 convolution like a fully connected layer?
So we have a feature map with dims (C, H, W), where C = (number of channels), H = height, W = width. I'll call positions in (H, W) "pixels". A 1x1 convolution will consist of C' (number of output channels of the convolution) kernels each with shape (C, 1, 1). So if we consider any pixel in the input feature map, we can apply a single (C, 1, 1) kernel to it to produce a (1, 1, 1) output. Applying C' different kernels will result in a (C', 1, 1) output. This is equivalent to applying a single fully connected layer to one pixel of the input feature map. Have a look at the following diagram to understand the action of a 1x1 convolution to a single pixel of the input feature map
The different colors represent different kernels of the convolution, corresponding to different output channels. You can see now how the kernels effectively comprise the weights of a single fully connected layer.
What is cross channel parametric pooling?
This is where I'm going to make a guess I'm 90% certain of (not 100% because I didn't read the paper). This is just an extension of the logic above, to whole feature maps rather than individual pixels. You're applying a cross-channel aggregation mechanism. The mechanism is parametric because it's not just a simple mean or sum or max, it's actually a parameterised weighted sum. Also note that the weights are held constant across all pixels (remember, that's how convolution kernels work). So it's essentially the same as applying the weights of a single fully connected layer to channels of a feature map in order to produce a different set of feature maps. But instead of applying the weights to individual neurons, you are applying them to the all the neurons of the feature map at the same time:
Link to paper
I'm trying to understand the region proposal network in faster rcnn. I understand what it's doing, but I still don't understand how training exactly works, especially the details.
Let's assume we're using VGG16's last layer with shape 14x14x512 (before maxpool and with 228x228 images) and k=9 different anchors. At inference time I want to predict 9*2 class labels and 9*4 bounding box coordinates. My intermediate layer is a 512 dimensional vector.
(image shows 256 from ZF network)
In the paper they write
"we randomly sample 256 anchors in an image to compute the loss
function of a mini-batch, where the sampled positive and negative
anchors have a ratio of up to 1:1"
That's the part I'm not sure about. Does this mean that for each one of the 9(k) anchor types the particular classifier and regressor are trained with minibatches that only contain positive and negative anchors of that type?
Such that I basically train k different networks with shared weights in the intermediate layer? Therefore each minibatch would consist of the training data x=the 3x3x512 sliding window of the conv feature map and y=the ground truth for that specific anchor type.
And at inference time I put them all together.
I appreciate your help.
Not exactly. From what I understand, the RPN predicts WHk bounding boxes per feature map, and then 256 are randomly sampled per the 1:1 criteria, and these are used as part of the computation for the loss function of that particular mini-batch. You're still only training one network, not k, since the 256 random samples are not of any particular type.
Disclaimer: I only started learning about CNNs a month ago, so I may not understand what I think I understand.
After going through the Caffe tutorial here: http://caffe.berkeleyvision.org/gathered/examples/mnist.html
I am really confused about the different (and efficient) model using in this tutorial, which is defined here: https://github.com/BVLC/caffe/blob/master/examples/mnist/lenet_train_test.prototxt
As I understand, Convolutional layer in Caffe simply calculate the sum of Wx+b for each input, without applying any activation function. If we would like to add the activation function, we should add another layer immediately below that convolutional layer, like Sigmoid, Tanh, or Relu layer. Any paper/tutorial I read on the internet applies the activation function to the neuron units.
It leaves me a big question mark as we only can see the Convolutional layers and Pooling layers interleaving in the model. I hope someone can give me an explanation.
As a site note, another doubt for me is the max_iter in this solver:
https://github.com/BVLC/caffe/blob/master/examples/mnist/lenet_solver.prototxt
We have 60.000 images for training, 10.000 images for testing. So why does the max_iter here only 10.000 (and it still can get > 99% accuracy rate)? What does Caffe do in each iteration?
Actually, I'm not so sure if the accuracy rate is the total correct prediction/test size.
I'm very amazed of this example, as I haven't found any example, framework that can achieve this high accuracy rate in that very short time (only 5 mins to get >99% accuracy rate). Hence, I doubt there should be something I misunderstood.
Thanks.
Caffe uses batch processing. The max_iter is 10,000 because the batch_size is 64. No of epochs = (batch_size x max_iter)/No of train samples. So the number of epochs is nearly 10. The accuracy is calculated on the test data. And yes, the accuracy of the model is indeed >99% as the dataset is not very complicated.
For your question about the missing activation layers, you are correct. The model in the tutorial is missing activation layers. This seems to be an oversight of the tutorial. For the real LeNet-5 model, there should be activation functions following the convolution layers. For MNIST, the model still works surprisingly well without the additional activation layers.
For reference, in Le Cun's 2001 paper, it states:
As in classical neural networks, units in layers up to F6 compute a dot product between their input vector and their weight vector, to which a bias is added. This weighted sum, denoted a_i, for unit i, is then passed through a sigmoid squashing function to produce the state of unit i ...
F6 is the "blob" between the two fully connected layers. Hence the first fully connected layers should have an activation function applied (the tutorial uses ReLU activation functions instead of sigmoid).
MNIST is the hello world example for neural networks. It is very simple to today's standard. A single fully connected layer can solve the problem with accuracy of about 92%. Lenet-5 is a big improvement over this example.
I'm new to caffe, in the MNIST example, I was thought that label should compared with softmax layers, but it was not the situation in lenet.prototxt.
I wonder why use InnerProduct result and label to get the accuracy, it seems unreasonable. Was it because I missed something in the layers?
The output dimension of the last inner product layer is 10, which corresponds to the number of classes (digits 0~9) of your problem.
The loss layer, takes two blobs, the first one being the prediction(ip2) and the second one being the label provided by the data layer.
loss_layer = SoftmaxLossLayer(name="loss", bottoms=[:ip2,:label])
It does not produce any outputs - all it does is to compute the loss function value, report it when backpropagation starts, and initiates the gradient with respect to ip2. This is where all magic starts.
After training ( in TEST phase), In the last layer desire results come from multiply weights and ip1( that are computed in last layer ); And each of class( one of 10 neurons) has max value is choosen.