# Questions tagged [convolution]

Convolution is a function-valued operation on two functions $f$ and $g$: $\int _{-\infty }^{\infty }f(\tau )g(t-\tau )d\tau$. Often used for obtaining the density of a sum of independent random variables. This tag should also be used for the inverse operation of deconvolution. DO NOT use this tag for convolutional neural networks.

452 questions
Filter by
Sorted by
Tagged with
13 views

### Is it acceptable to use pooling layers in variational autoencoders?

When training a model for image classification it is common to use pooling layers to reduce the dimensionality, as we only care about the final node values corresponding to the categorical ...
• 13
12 views

### what are the differences between receptive field (RF) and field-of-views (FOV) in DeepLab papers?

I am learning the deeplab models. However, some concepts in the papers made me confused. Receptive field (RF) and field-of-views (FOV) are two concepts mentioned in the Deeplabv1 paper. I know that ...
• 111
333 views

### Density of $|t_1 - t_2|$ where $t_1$ and $t_2$ are iid with $P(t) = \alpha e^{-t\alpha}$

I am trying to answer the following question from my quantum mechanics textbook and my probability theory is admittedly rusty (this is not schoolwork as should be clear from my post history on Phys ...
• 203
1 vote
48 views

### Order Statistics - Percentile Range of Normal Mixture of Normals

Say I have draw N values from a normal distribution [$\mu_1$, $\sigma_1$]. Below are 10 sampled points compared to the normal distribution they're sampled from I then create a normal mixture of ...
13 views

### How to calculate the maximum and minimum number of feature maps or filters in a CNN?

I programmed an Autoencoder to encode images, and I want to know what is the max and min number of feature maps in layers.Conv2D(?, (3, 3), activation='relu', padding='same')(input_img). The size of ...
92 views

### Distribution closed under convolution and truncation followed by convolution

Let $D(\theta)$ denote an absolutely continuous distribution on $\mathbb{R}$. (The finite dimensional vector $\theta$ collects the parameters of the distribution.) Assume that the p.d.f. of $D(\theta)$...
• 495
1 vote
67 views

### How can I build a 6D Convolution Layer in a CNN, using R or Python? [closed]

As stated in the title, I would like to build an N-dimensional Convolutional Layer as part of a Convolution Neural Network, without doing dimensionality reduction on my data; because I have multiple ...
• 11
15 views

### Unclear Architecture of MNIST Neural Network

I am trying to reproduce a Neural Network trained to detect whether there is a 0-3 digit in an image with another confounding image. The paper I am following lists the corresponding architecture: A ...
32 views

### What "Convolution filters along the time axis" means?

Suppose that I have a tensor of height:25 and width:50. Height is my temporal axis, therefore I have a window of 25 time steps. Therefore my input tensor is: I want to extract temporal features / ...
• 151
141 views

### Calculating the receptive field of a CNN when stride > kernelsize

I am reading about how to calculate the receptive field on Araujo, et al., "Computing Receptive Fields of Convolutional Neural Networks", Distill, 2019. however there is one thing that ...
1 vote
15 views

### Relationship between convolutions in neural nets and probabilistic convolution

In probability theory, convolution extends to produce the distribution of the sum of two independent random variables. I've only ever seen this in the context of univariate random variables. However, ...
• 249
1 vote
6 views

### Are Class Activated Maps an example of backpropagated Explainable AI?

Class Activated Maps (NOT GradCAM) is listed in van der Velden et al as a back-propagation approach in table 1, but that is not actually true, is it? CAM doesn't use back-propagation in the method, ...
1 vote
21 views

### How to calculate how much each error source contributes to an overall accuracy metric?

I am building a model of overall accuracy for a robotic system, I have various error sources from assembly, calibrations, measurements, imaging, for each error I have a PDF of the error it induces on ...
37 views

### How to average multiple non-normal distributions?

I have the following statistics of two independent random variables: First random variable: ...
• 51
12 views

### Discovering alternative waveforms for Fourier transform (alternatives to sinusoidal waves)

I'm working on a classification algorithm that utilises a discrete Fourier transform on sensor data, as a means to detect the presence of an oscillating signal. My issue is that the signal-generating ...
1 vote
25 views

### Mean conditional on sum

Let $X$ and $Y$ be two random variables. I am interested in how $\mathrm{E}[X | X + Y = z]$ changes as $z$ changes. Intuitively, if $X$ and $Y$ are independent, the conditional mean should be ...
342 views

### How to determine the no of multiplication operations in convolution operation?

Let's say We have an input of size 28×28×192. We apply 32, 5×5 convolution filters with padding "same". How many multiplication operations will be there in total? I know there will be ...
52 views

### How to take 1000 samples from distribution X and then use MLE to prove they came from distribution X?

I am trying to do: find 1000 points that represent samples from distribution X with parameters $(a,b,c,\ldots, d)$ be guaranteed that the MLEs for those 1000 points are $(a,b,c,\ldots, d)$ with $99\%$...
10 views

### What is difference between the joint probability distribution and the sum/convolution, of 2 dists? [duplicate]

Google is coming up a bit short when I searched for "joint vs sum random variables". Perhaps someone can provide an authoritative answer to compare and contrast the sum/convolution of 2 ...
968 views

### Why does a 1D convolution increase the size of the output, while a 2D convolution tends to decrease (such as in a CNN?)

The function np.convolve is a 1D convolution (e.g. when both inputs are 1D). It results in a larger output size. ...
• 205
43 views

• 171
62 views

### Sparsity of connections in a CNN - why is it beneficial?

In his CNN video, Andrew Ng explains the benefits of convolutions over FC, and he mentions sparsity of connections as being one of them. Can someone please explain what is the benefit of sparsity when ...
• 183
1 vote
82 views

### How to explain the high accuracy and F1 score on the test set with a huge binary crossentropy loss?

I'll provide a little of introduction based on my example. I have a small collection of RGB (but 'gray-looking') brain MRI photos, divided into 2 classes: healthy and tumor. My data split looks like ...
18 views

### does kernel or filter update through backpropagation , is require fully connected layers?

does CNN itself without FC has the ability to make backpropagation to update its filters, after comparing them with the output and calculating the loss, and then pass them through the FCC?
1 vote
157 views

### GAN artifacts on borders

not quite a math-question, but I have a doubt. I'm trying to build from scratch the Pix2pix network, on the facades dataset, and I think I finally got a good model (from the paper I borrowed just the ...
1 vote
59 views

### Why not using Convolutional pooling instead of MaxPooling to avoid invariance

I was reading the paper of [Geoffrey Hinton: Capsule network], and I watch it's talk on Youtube about the problem of Conv Network is actually the (max) pooling layer, since we don't want to be ...
1 vote
47 views

### How does fix-up initialization avoid prevent the neurons from updating in the exact same direction?

We know zero initialization is bad: Pitfall: all zero initialization. Lets start with what we should not do. Note that we do not know what the final value of every weight should be in the trained ...
• 237
93 views

### Validation loss is stuck at certain loss [duplicate]

I am trying to implement 1D-CNN to time-series data. Here is my data. Here is my dataframe: I am trying to forecast each of these futures. I am following the TensorFlow forecasting guide. Based on ...
• 245