I am trying pixelcnn, which is auto-regressive generative model. After training, the model receive an all-zero tensor and generate the next pixel form the left top coner. Now that the model parameters are fixed, does the model only can produce the same outputs starting from the same zero tensor? How to produce different samples?
Yes, you always provide an all-zero tensor. However, for PixelCNN each pixel location is represented by a distribution. So when you do the forward pass you then sample from a random distribution at the end. That is how the pixel values are different each run.
This is of course because PixelCNN is a probabilistic neural network. So the pixels, as mentioned before, are all represented by conditional probability distributions of all the layers below, not just point estimates.
Related
I gather that Masking layers in Keras are commonly used for handling data inputs with varying timesteps. Based on the documentation, I understand that if all of the features for a given timestep equal the mask value, then that timestep will be skipped in downstream layers.
For my problem, I am instead interested in using masking for features, where the data input shape to the network is (batch_size, num_timesteps, num_features). Essentially, I want to be able to predict a timeseries one step into the future with num_features features, but assuming that I won't always have all the features from the previous timestep to base my prediction on.
For example, one could predict RGB values one timestep into the future for a pixel in a video stream based on partial data from a previous timestep. At every timestep the output should be all RGB, but some timesteps you may get only RG, or only RB, or only BG, but you never know which partial data you'll have at each timestep to make your prediction. This is why I want to somehow be able to indicate a feature as masked during training to accommodate this kind of prediction.
It may be that Masking in Keras is not the correct mechanism to achieve this. What is the correct type of network layer that would give me this behavior?
I was trying to visualize the output of all the activation functions in each layers of my fully-connected network and I was surprised when I checked the last layer. It is a very simple regression model and the output layer has therefore one neuron. I know it would be better to visualize it as a scalar, but I was just trying to visualize it using histograms and I realized that the value is somehow split. Does this make any sense? I would rather expect that it cannot be visualized at all or that the histogram would consist of a single point.
I'm trying to predict sequences of 2D coordinates. But I don't want only the most probable future path but all the most probable paths to visualize it in a grid map.
For this I have traning data consisting of 40000 sequences. Each sequence consists of 10 2D coordinate pairs as input and 6 2D coordinate pairs as labels.
All the coordinates are in a fixed value range.
What would be my first step to predict all the probable paths? To get all probable paths I have to apply a softmax in the end, where each cell in the grid is one class right? But how to process the data to reflect this grid like structure? Any ideas?
A softmax activation won't do the trick I'm afraid; if you have an infinite number of combinations, or even a finite number of combinations that do not already appear in your data, there is no way to turn this into a multi-class classification problem (or if you do, you'll have loss of generality).
The only way forward I can think of is a recurrent model employing variational encoding. To begin with, you have a lot of annotated data, which is good news; a recurrent network fed with a sequence X (10,2,) will definitely be able to predict a sequence Y (6,2,). But since you want not just one but rather all probable sequences, this won't suffice. Your implicit assumption here is that there is some probability space hidden behind your sequences, which affects how they play out over time; so to model the sequences properly, you need to model that latent probability space. A Variational Auto-Encoder (VAE) does just that; it learns the latent space, so that during inference the output prediction depends on sampling over that latent space. Multiple predictions over the same input can then result in different outputs, meaning that you can finally sample your predictions to empirically approximate the distribution of potential outputs.
Unfortunately, VAEs can't really be explained within a single paragraph over stackoverflow, and even if they could I wouldn't be the most qualified person to attempt it. Try searching the web for LSTM-VAE and arm yourself with patience; you'll probably need to do some studying but it's definitely worth it. It might also be a good idea to look into Pyro or Edward, which are probabilistic network libraries for python, better suited to the task at hand than Keras.
I have a large data set (~30 million data-points with 5 features) that I have reduced using K-means down to 200,000 clusters. The data is a time-series with ~150,000 time-steps. The data on which I would like to train the model is the presence of particular clusters at each time-step. The purpose of the predictive model is generate a generalized sequence similar to generating syntactically correct sentences from a model trained on word sequences. The easiest way to think about this data is that I'm trying to predict the pixels in the next video frame from pixels in the current video frame in order to generate a new sequence of frames that approximate the original sequence.
The raw and sparse representation at each time-step would be 200,000 binary values representing which clusters are present or not at that time step. Note, no more than 200 clusters may be present in any one time-step and thus this representation is extremely sparse.
What is the best representation to convert this sparse vector to a dense vector that would be more suitable to time-series prediction using Tensorflow?
I initially had in mind a RNN / LSTM trained on the vectors at each time-step, but due to the size of the training vector I'm now wondering if a convolution approach would be more suitable.
Note, I have not actually used tensorflow beyond some simple tutorials, but have have previously used OpenCV ML functions. Please consider me a novice in your responses.
Thank you.
There is a type of architecture that I would like to experiment with in TensorFlow.
The idea is to compose 2-D filter kernels by a combination of 1-D filters.
From the paper:
Simplifying ConvNets through Filter Compositions
The essence of our proposal consists of decomposing the ND kernels of a traditional network into N consecutive layers of 1D kernels.
...
We propose DecomposeMe which is an architecture consisting of decomposed layers. Each decomposed layer represents a N-D convolutional layer as a composition of 1D filters and, in addition, by including a non-linearity
φ(·) in-between.
...
Converting existing structures to decomposed ones is a straight forward process as
each existing ND convolutional layer can systematically be decomposed into sets of
consecutive layers consisting of 1D linearly rectified kernels and 1D transposed kernels
as shown in Figure 1.
If I understand correctly, a single 2-D convolutional layer is replaced with two consecutive 1-D convolutions?
Considering that the weights are shared and transposed, it is not clear to me how exactly to implement this in TensorFlow.
I know this question is old and you probably already figured it out, but it might help someone else with the same problem.
Separable convolution can be implemented in tensorflow as follows (details omitted):
X= placeholder(float32, shape=[None,100,100,3]);
v1=Variable(truncated_normal([d,1,3,K],stddev=0.001));
h1=Variable(truncated_normal([1,d,K,N],stddev=0.001));
M1=relu(conv2(conv2(X,v1),h1));
Standard 2d convolution with a column vector is the same as convolving each column of the input with that vector. Convolution with v1 produces K feature maps (or an output image with K channels), which is then passed on to be convolved by h1 producing the final desired number of featuremaps N.
Weight sharing, according to my knowledge, is simply a a misleading term, which is meant to emphasize the fact that you use one filter that is convolved with each patch in the image. Obviously you're going to use the same filter to obtain the results for each output pixel, which is how everyone does it in image/signal processing.
Then in order to "decompose" a convolution layer as shown on page 5, it can be done by simply adding activation units in between the convolutions (ignoring biases):
M1=relu(conv2(relu(conv2(X,v1)),h1));
Not that each filter in v1 is a column vector [d,1], and each h1 is a row vector [1,d]. The paper is a little vague, but when performing separable convolution, this is how it's done. That is, you convolve the image with the column vectors, then you convolve the result with the horizontal vectors, obtaining the final result.