New answers tagged

0

In machine learning, there's often no clear "why" or "reason" we can point to. These are different methods, which implicitly embed different biases or different priors or different assumptions, and thus will work better in different situations.


1

The order of dimensions usually depends on the library you are using Tensorflow : (Height,Width,Channel) : HWC PyTorch: (Channel,Height,Width) : CHW In your case the image seems to go with CHW , since you cant apply 5x5 filters on a 1x128 image As for the final shape , we follow the rule : for input length N, and k-size filter , output-length = N-k+1 so each ...


0

The image dimensions can either be CHW (Channel, Height, Width) or HWC (Height, Width, Channel). In this case, the image dimensions appear to be CHW, and so you're possibly looking to apply a 5x5 kernel to a 28x28 image.


0

Am I missing something about this problem that renders it particularly immune to the application of mature technologies to hide data representation from the abstract functions? I would say so. And that is empirical experience of what works. In theory data is data. A general artificial intelligence agent should be able to classify a picture as a cat or a dog ...


0

Having a powerful machine learning chip in your device allows you to do things that you probably have not thought of. Things like improving the images that your camera is taking. Cutting out the background and replacing it with a boring beige background, very valuable if you are in a video conference and want to show your face to people but not your home. Or ...


2

This is likely to be a challenging task, and if you're not familiar with image classification, I suspect it may be beyond what you can reasonably do right now. I would suggest you investigate classical image processing techniques. You might try applying morphological operators, such as closing (dilation followed by erosion) or opening followed by ...


1

Given a new point $x_q$, we find the $k$ closest points $x_1,\ldots,x_k$. Each one of them is associated with a cluster $f(x_1),\ldots,f(x_k)$. We assign $x_q$ to the most common cluster among $f(x_1),\ldots,f(x_k)$. The argmax equation is just a fancy way of expressing this idea: for any cluster $v$, the quantity $\sum_{i=1}^k \delta(v,f(x_i))$ is just the ...


0

A single-layer perceptron can only learn linearly separable patterns (with 100% accuracy), as noted in the Wikipedia article on perceptrons. (Advice for the future: check standard references before asking here.) A multilayer feedforward neural network (sometimes called multilayer perceptron) can learn other patterns. If someone just says "perceptron&...


Top 50 recent answers are included