Related
Background:
Thanks for your attention! I am learning the basic knowledge of 2D convolution, linear algebra and PyTorch. I encounter the implementation problem about the psedo-inverse of the convolution operator. Specifically, I have no idea about how to implement it in an efficient way. Please see the following problem statements for details. Any help/tip/suggestion is welcomed.
(Thanks a lot for your attention!)
The Original Problem:
I have an image feature x with shape [b,c,h,w] and a 3x3 convolutional kernel K with shape [c,c,3,3]. There is y = K * x. How to implement the corresponding pseudo-inverse on y in an efficient way?
There is [y = K * x = Ax], how to implement [x_hat = (A^+)y]?
I guess that there should be some operations using torch.fft. However, I still have no idea about how to implement it. I do not know if there exists an implementation previously.
import torch
import torch.nn.functional as F
c = 32
K = torch.randn(c, c, 3, 3)
x = torch.randn(1, c, 128, 128)
y = F.conv2d(x, K, padding=1)
print(y.shape)
# How to implement pseudo-inverse for y = K * x in an efficient way?
Some of My Efforts:
I may know that the 2D convolution is a linear operator. It is equivalent to a "matrix product" operator. We can actually write out the matrix form of the convolution and calculate its psedo-inverse. However, I think this type of operation will be inefficient. And I have no idea about how to implement it in an efficient way.
According to Wikipedia, the psedo-inverse may satisfy the property of A(A_pinv(x))=x, where A is the convolutional operator, A_pinv is its psedo-inverse, and x may be any image feature.
(Thanks again for reading such a long post!)
This takes the problem to another level.
The convolution itself is a linear operation, you can determine the matrix of the operation and solve a least square problem directly [1], or compute the pseudo-inverse as you mentioned, and then apply to different outputs and predicting a projection of the input.
I am changing your code to using padding=0
import torch
import torch.nn.functional as F
# your code
c = 32
K = torch.randn(c, c, 1, 1)
x = torch.randn(4, c, 128, 128)
y = F.conv2d(x, K, bias=torch.zeros((c,)))
Also, as you probably already suggested the convolution can be computed as ifft(fft(h)*fft(x)). However, the conv2d function is a cross-correlation, so you have to conjugate the filter leading to ifft(fft(h)*fft(x)), also you have to apply this to two axes, and you have to make sure the FFT is calcuated using the same representation (size), since the data is real, we can apply multi-dimensional real FFT. To be complete, conv2d works on multiple channels, so we have to calculate summations of convolutions. Since the FFT is linear, we can simply compute the summations on the frequency domain
using einsum.
s = y.shape[-2:]
K_f = torch.fft.rfftn(K, s)
x_f = torch.fft.rfftn(x, s)
y_f = torch.einsum('jkxy,ikxy->ijxy', K_f.conj(), x_f)
y_hat = torch.fft.irfftn(y_f, s)
Except for the borders it should be accurate (remember FFT computes a cyclic convolution).
torch.max(abs(y_hat[:,:,:-2,:-2] - y[:,:,:,:]))
Now, notice the pattern jk,ik->ij on the einsum, that means y_f[i,j] = sum(K_f[j,k] * x_f[i,k]) = x_f # K_f.T, if # is the matrix product on the first two dimensions. So to invert this operation we have to can interpret the first two dimensions as matrices. The function pinv will compute pseudo-inverses on the last two axes, so in order to use that we have to permute the axes. If we right multiply the output by the pseudo-inverse of transposed K_f we should invert this operation.
s = 128,128
K_f = torch.fft.rfftn(K, s)
K_f_inv = torch.linalg.pinv(K_f.T).T
y_f = torch.fft.rfftn(y_hat, s)
x_f = torch.einsum('jkxy,ikxy->ijxy', K_f_inv.conj(), y_f)
x_hat = torch.fft.irfftn(x_f, s)
print(torch.mean((x - x_hat)**2) / torch.mean((x)**2))
Nottice that I am using the full convolution, but the conv2d actually cropped the images. Let's apply that
y_hat[:,:,128-(k-1):,:] = 0
y_hat[:,:,:,128-(k-1):] = 0
Repeating the calculation you will see that the input is not accurate anymore, so you have to be careful about what you do with your convolution, but in some situations where you can get this to work it will be in fact efficient.
s = 128,128
K_f = torch.fft.rfftn(K, s)
K_f_inv = torch.linalg.pinv(K_f.T).T
y_f = torch.fft.rfftn(y_hat, s)
x_f = torch.einsum('jkxy,ikxy->ijxy', K_f_inv.conj(), y_f)
x_hat = torch.fft.irfftn(x_f, s)
print(torch.mean((x - x_hat)**2) / torch.mean((x)**2))
I am trying to compute a weighted output from multiple parallel models using Keras' Merge layer. I'm using Theano backend.
I have L parallel models (Ci). Each of their output layer is a k-sized softmax.
There is one model (N), its output is a L-sized softmax.
Here is what I have so far:
Parallel models (Ci) each with k dimension in the output layer:
model.add(Dense(K, activation='softmax', W_regularizer=l2(0.001),init='normal'))
The weighing model (N), output layer:
model.add(Dense(L, activation='softmax', W_regularizer=l2(0.001), init='normal'))
The merger is as follows:
model.add(Merge(layers=model_group,
mode=lambda model_group: self.merge_fun(model_group, L),
output_shape = (None, k)))
where "model_group" is a (L+1)-length list [N, C1, C2, ..., CL], and merge_fun's signature is:
def merge_fun(self, model_group, L):
Mathematically, I would like the output of the merged layer to be a weighted sum:
out = N[1]x([C11, C12, C13, .., C1k]) + N[2]x([C21, C22, C23, ..., C2k]) + ... + N[L]x([CL1, CL2, CL3, ..., CLk]),
where out is a vector of size k
How can I use the Merge layer to achieve this ?
I know that the magic would probably have to happen in the 'merge_fun', but I am not sure how to perform matrix algebra in Keras. The tensor parameters don't have a "shape" parameter - they have a keras_shape = (None, K or L) - but I am not sure how to combine parallel models' output into a matrix.
I tried using a local evaluation of the following expressions:
K.concatenate([model_group[1], model_group[2]], axis=0)*model_group[0]
and
model_group[0] * K.concatenate([model_group[1], model_group[2]], axis=0)
both of which didn't throw an error, so I can't use this as a guide. After the multiplication, the result returned did not have the keras_shape variable, so I'm not sure what the shape of the result is.
Any suggestions ?
What I advise you is to use a functional API and use this is in a following manner:
Define the L output models:
softmax_1 = Dense(K, activation='softmax', ...))(input_to_softmax_1)
softmax_2 = Dense(K, activation='softmax', ...))(input_to_softmax_2)
...
softmax_L = Dense(K, activation='softmax', ...))(input_to_softmax_L)
Define the merge softmax:
merge_softmax= Dense(L, activation='softmax', ...)(input_to_merge_softmax)
merge_softmax = Reshape((1, L))(merge_softmax)
Merge and reshape the bag of L models:
bag_of_models = merge([softmax_1, ..., softmax_L], mode = 'concat')
bag_of_models = Reshape((L, K))(bag_of_models)
Compute the final merged softmax:
final_result = merge([bag_of_models, merge_softmax], mode = 'dot', dot_axes = [1, 2])
final_result = Reshape((K, ))(final_result)
Of course - depending on your topology - different tensor might be the same (e.g. input to different softmaxes). I tested this on my machine but due to extensive refactoring - I might made mistake - so if you fin one - please inform me.
The solution with Sequential is much less clear and a little bit cumbersome - but if you want one - please write in the comment so I will update my answer.
Could someone give a clear explanation of backpropagation for LSTM RNNs?
This is the type structure I am working with. My question is not posed at what is back propagation, I understand it is a reverse order method of calculating the error of the hypothesis and output used for adjusting the weights of neural networks. My question is how LSTM backpropagation is different then regular neural networks.
I am unsure of how to find the initial error of each gates. Do you use the first error (calculated by hypothesis minus output) for each gate? Or do you adjust the error for each gate through some calculation? I am unsure how the cell state plays a role in the backprop of LSTMs if it does at all. I have looked thoroughly for a good source for LSTMs but have yet to find any.
That's a good question. You certainly should take a look at suggested posts for details, but a complete example here would be helpful too.
RNN Backpropagaion
I think it makes sense to talk about an ordinary RNN first (because LSTM diagram is particularly confusing) and understand its backpropagation.
When it comes to backpropagation, the key idea is network unrolling, which is way to transform the recursion in RNN into a feed-forward sequence (like on the picture above). Note that abstract RNN is eternal (can be arbitrarily large), but each particular implementation is limited because the memory is limited. As a result, the unrolled network really is a long feed-forward network, with few complications, e.g. the weights in different layers are shared.
Let's take a look at a classic example, char-rnn by Andrej Karpathy. Here each RNN cell produces two outputs h[t] (the state which is fed into the next cell) and y[t] (the output on this step) by the following formulas, where Wxh, Whh and Why are the shared parameters:
In the code, it's simply three matrices and two bias vectors:
# model parameters
Wxh = np.random.randn(hidden_size, vocab_size)*0.01 # input to hidden
Whh = np.random.randn(hidden_size, hidden_size)*0.01 # hidden to hidden
Why = np.random.randn(vocab_size, hidden_size)*0.01 # hidden to output
bh = np.zeros((hidden_size, 1)) # hidden bias
by = np.zeros((vocab_size, 1)) # output bias
The forward pass is pretty straightforward, this example uses softmax and cross-entropy loss. Note each iteration uses the same W* and h* arrays, but the output and hidden state are different:
# forward pass
for t in xrange(len(inputs)):
xs[t] = np.zeros((vocab_size,1)) # encode in 1-of-k representation
xs[t][inputs[t]] = 1
hs[t] = np.tanh(np.dot(Wxh, xs[t]) + np.dot(Whh, hs[t-1]) + bh) # hidden state
ys[t] = np.dot(Why, hs[t]) + by # unnormalized log probabilities for next chars
ps[t] = np.exp(ys[t]) / np.sum(np.exp(ys[t])) # probabilities for next chars
loss += -np.log(ps[t][targets[t],0]) # softmax (cross-entropy loss)
Now, the backward pass is performed exactly as if it was a feed-forward network, but the gradient of W* and h* arrays accumulates the gradients in all cells:
for t in reversed(xrange(len(inputs))):
dy = np.copy(ps[t])
dy[targets[t]] -= 1
dWhy += np.dot(dy, hs[t].T)
dby += dy
dh = np.dot(Why.T, dy) + dhnext # backprop into h
dhraw = (1 - hs[t] * hs[t]) * dh # backprop through tanh nonlinearity
dbh += dhraw
dWxh += np.dot(dhraw, xs[t].T)
dWhh += np.dot(dhraw, hs[t-1].T)
dhnext = np.dot(Whh.T, dhraw)
Both passes above are done in chunks of size len(inputs), which corresponds to the size of the unrolled RNN. You might want to make it bigger to capture longer dependencies in the input, but you pay for it by storing all outputs and gradients per each cell.
What's different in LSTMs
LSTM picture and formulas look intimidating, but once you coded plain vanilla RNN, the implementation of LSTM is pretty much same. For example, here is the backward pass:
# Loop over all cells, like before
d_h_next_t = np.zeros((N, H))
d_c_next_t = np.zeros((N, H))
for t in reversed(xrange(T)):
d_x_t, d_h_prev_t, d_c_prev_t, d_Wx_t, d_Wh_t, d_b_t = lstm_step_backward(d_h_next_t + d_h[:,t,:], d_c_next_t, cache[t])
d_c_next_t = d_c_prev_t
d_h_next_t = d_h_prev_t
d_x[:,t,:] = d_x_t
d_h0 = d_h_prev_t
d_Wx += d_Wx_t
d_Wh += d_Wh_t
d_b += d_b_t
# The step in each cell
# Captures all LSTM complexity in few formulas.
def lstm_step_backward(d_next_h, d_next_c, cache):
"""
Backward pass for a single timestep of an LSTM.
Inputs:
- dnext_h: Gradients of next hidden state, of shape (N, H)
- dnext_c: Gradients of next cell state, of shape (N, H)
- cache: Values from the forward pass
Returns a tuple of:
- dx: Gradient of input data, of shape (N, D)
- dprev_h: Gradient of previous hidden state, of shape (N, H)
- dprev_c: Gradient of previous cell state, of shape (N, H)
- dWx: Gradient of input-to-hidden weights, of shape (D, 4H)
- dWh: Gradient of hidden-to-hidden weights, of shape (H, 4H)
- db: Gradient of biases, of shape (4H,)
"""
x, prev_h, prev_c, Wx, Wh, a, i, f, o, g, next_c, z, next_h = cache
d_z = o * d_next_h
d_o = z * d_next_h
d_next_c += (1 - z * z) * d_z
d_f = d_next_c * prev_c
d_prev_c = d_next_c * f
d_i = d_next_c * g
d_g = d_next_c * i
d_a_g = (1 - g * g) * d_g
d_a_o = o * (1 - o) * d_o
d_a_f = f * (1 - f) * d_f
d_a_i = i * (1 - i) * d_i
d_a = np.concatenate((d_a_i, d_a_f, d_a_o, d_a_g), axis=1)
d_prev_h = d_a.dot(Wh.T)
d_Wh = prev_h.T.dot(d_a)
d_x = d_a.dot(Wx.T)
d_Wx = x.T.dot(d_a)
d_b = np.sum(d_a, axis=0)
return d_x, d_prev_h, d_prev_c, d_Wx, d_Wh, d_b
Summary
Now, back to your questions.
My question is how is LSTM backpropagation different then regular Neural Networks
The are shared weights in different layers, and few more additional variables (states) that you need to pay attention to. Other than this, no difference at all.
Do you use the first error (calculated by hypothesis minus output) for each gate? Or do you adjust the error for each gate through some calculation?
First up, the loss function is not necessarily L2. In the example above it's a cross-entropy loss, so initial error signal gets its gradient:
# remember that ps is the probability distribution from the forward pass
dy = np.copy(ps[t])
dy[targets[t]] -= 1
Note that it's the same error signal as in ordinary feed-forward neural network. If you use L2 loss, the signal indeed equals to ground-truth minus actual output.
In case of LSTM, it's slightly more complicated: d_next_h = d_h_next_t + d_h[:,t,:], where d_h is the upstream gradient the loss function, which means that error signal of each cell gets accumulated. But once again, if you unroll LSTM, you'll see a direct correspondence with the network wiring.
I think your questions could not be answered in a short response. Nico's simple LSTM has a link to a great paper from Lipton et.al., please read this. Also his simple python code sample helps to answer most of your questions.
If you understand Nico's last sentence
ds = self.state.o * top_diff_h + top_diff_s
in detail, please give me a feed back. At the moment I have a final problem with his "Putting all this s and h derivations together".
I have a convolutional neural network whose output is a 4-channel 2D image. I want to apply sigmoid activation function to the first two channels and then use BCECriterion to computer the loss of the produced images with the ground truth ones. I want to apply squared loss function to the last two channels and finally computer the gradients and do backprop. I would also like to multiply the cost of the squared loss for each of the two last channels by a desired scalar.
So the cost has the following form:
cost = crossEntropyCh[{1, 2}] + l1 * squaredLossCh_3 + l2 * squaredLossCh_4
The way I'm thinking about doing this is as follow:
criterion1 = nn.BCECriterion()
criterion2 = nn.MSECriterion()
error = criterion1:forward(model.output[{{}, {1, 2}}], groundTruth1) + l1 * criterion2:forward(model.output[{{}, {3}}], groundTruth2) + l2 * criterion2:forward(model.output[{{}, {4}}], groundTruth3)
However, I don't think this is the correct way of doing it since I will have to do 3 separate backprop steps, one for each of the cost terms. So I wonder, can anyone give me a better solution to do this in Torch?
SplitTable and ParallelCriterion might be helpful for your problem.
Your current output layer is followed by nn.SplitTable that splits your output channels and converts your output tensor into a table. You can also combine different functions by using ParallelCriterion so that each criterion is applied on the corresponding entry of output table.
For details, I suggest you read documentation of Torch about tables.
After comments, I added the following code segment solving the original question.
M = 100
C = 4
H = 64
W = 64
dataIn = torch.rand(M, C, H, W)
layerOfTables = nn.Sequential()
-- Because SplitTable discards the dimension it is applied on, we insert
-- an additional dimension.
layerOfTables:add(nn.Reshape(M,C,1,H,W))
-- We want to split over the second dimension (i.e. channels).
layerOfTables:add(nn.SplitTable(2, 5))
-- We use ConcatTable in order to create paths accessing to the data for
-- numereous number of criterions. Each branch from the ConcatTable will
-- have access to the data (i.e. the output table).
criterionPath = nn.ConcatTable()
-- Starting from offset 1, NarrowTable will select 2 elements. Since you
-- want to use this portion as a 2 dimensional channel, we need to combine
-- then by using JoinTable. Without JoinTable, the output will be again a
-- table with 2 elements.
criterionPath:add(nn.Sequential():add(nn.NarrowTable(1, 2)):add(nn.JoinTable(2)))
-- SelectTable is simplified version of NarrowTable, and it fetches the desired element.
criterionPath:add(nn.SelectTable(3))
criterionPath:add(nn.SelectTable(4))
layerOfTables:add(criterionPath)
-- Here goes the criterion container. You can use this as if it is a regular
-- criterion function (Please see the examples on documentation page).
criterionContainer = nn.ParallelCriterion()
criterionContainer:add(nn.BCECriterion())
criterionContainer:add(nn.MSECriterion())
criterionContainer:add(nn.MSECriterion())
Since I used almost every possible table operation, it looks a little bit nasty. However, this is the only way I could solve this problem. I hope that it helps you and others suffering from the same problem. This is how the result looks like:
dataOut = layerOfTables:forward(dataIn)
print(dataOut)
{
1 : DoubleTensor - size: 100x2x64x64
2 : DoubleTensor - size: 100x1x64x64
3 : DoubleTensor - size: 100x1x64x64
}
I don't understand a passage in the article about the VGGNet. Maybe someone can help.
In my opinion, the number of weights in a convolutional layer is
p=w*h*d*n+n
where w is the width of the filters, h the height of the filters, d the depth of the filters and n the num of the filters.
In the article the following is written:
assuming that both the input and the output of a three-layer 3 × 3 onvolution stack has C channels, the stack is parametrised by 3*(3^2*C^2) = 27C^2
weights; at the same time, a single 7 × 7 conv. layer would require 7^2*C^2 = 49C^2 parameters.
I do not understand, what is meant by channels here, and why this formula is used.
Can someone explain this to me?
Thanks in advance.
Your intuition is correct; we just need to unpack their explanation a bit. For the first case:
w = 3 # filter width
h = 3 # filter height
d = C # filter depth (number of channels is same as number of input filters; eg RGB is C=3)
n = C # number of output filters/channels
This then makes whdn = 9C^2 parameters. Then, they also say there are three of these stacked, so thats 27C^2.
For a single 7x7 filter, then it's all the same 7x7xCxCx1.
The final difference is that you add n once more at the end in your original post; that is the bias terms, which in VGG they skip (many people skip bias terms; their value is debatable in some settings).