filter
In plain wordsIn an image-recognition network, a small grid of learned numbers that acts as a pattern detector, for vertical edges, a patch of fur or a wheel, scanned across the whole picture.
Convolutional networks
The small tensor of learned weights, also called a kernel, that a convolutionalconvolution, n.: An operation that slides a small grid of weights across an image or signal, multiplying and adding at each spot, to measure how strongly a pattern appears at every location. layer slides over its input. For a C_in-channel input and kernel size k × k, a filter has shape k × k × C_in, so it combines spatial and cross-channel information; its response at every position forms one feature mapfeature map, n.: The grid of numbers produced when one learned pattern detector is scanned across an image, showing where, and how strongly, that pattern appears., and a layer with C_out filters produces C_out output channels. First-layer filters of networks trained on natural images typically resemble oriented edge and color-blob detectors, similar to Gabor filters and to receptive fields in primary visual cortex, while deeper filters respond to textures, parts and objects, as revealed by feature visualizationfeature visualization, n.: Making pictures that show what a part of an image-recognition network is looking for, by repeatedly adjusting an image until that part responds as strongly as possible.. The term is borrowed from signal processing, where a linear filter is applied by convolution.
Visualizing the first-layer filters showed mostly edge detectors and a few color blobs.
Origin
Borrowed from signal and image processing, where hand-designed linear filters, such as smoothing and edge-detection kernels, were applied by convolution decades before deep learning. In 1989 a Bell Labs zip-code system by John Denker and colleagues still used hand-designed kernels; Yann LeCun and colleagues showed the same year that they could be learned by backpropagation.
Usage note
'Filter' and 'kernel' are often interchangeable, though some authors reserve 'kernel' for one 2-D slice of a filter. Not to be confused with the kernelkernel, n.: A word with several AI meanings: a similarity measure used by some learning methods; a small grid of numbers that scans an image inside a neural network; or a small program that runs on a GPU. of kernel methods, data quality filteringquality filtering, n.: Automatically sorting through huge piles of collected text and keeping only the parts judged useful for training, such as well-written articles, while discarding spam, gibberish and junk. or content filters.
Related
- convolutionconvolution, n.: An operation that slides a small grid of weights across an image or signal, multiplying and adding at each spot, to measure how strongly a pattern appears at every location.
- feature mapfeature map, n.: The grid of numbers produced when one learned pattern detector is scanned across an image, showing where, and how strongly, that pattern appears.
- convolutional neural networkconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects.
- receptive fieldreceptive field, n.: The patch of the original input, such as a region of an image, that a particular neuron inside a network can see and respond to.
- feature visualizationfeature visualization, n.: Making pictures that show what a part of an image-recognition network is looking for, by repeatedly adjusting an image until that part responds as strongly as possible.
- weight sharingweight sharing, n.: Reusing the same learned numbers in many places in a network, for example one edge detector applied at every spot in an image, so the network needs fewer parameters and learns from less data.
Mentioned in
convolutionconvolution, n.: An operation that slides a small grid of weights across an image or signal, multiplying and adding at each spot, to measure how strongly a pattern appears at every location., convolutional neural networkconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects., feature mapfeature map, n.: The grid of numbers produced when one learned pattern detector is scanned across an image, showing where, and how strongly, that pattern appears., kernelkernel, n.: A word with several AI meanings: a similarity measure used by some learning methods; a small grid of numbers that scans an image inside a neural network; or a small program that runs on a GPU., kernel trickkernel trick, n.: A mathematical shortcut that lets simple straight-line methods find curved patterns, by computing similarities between examples as if the data had been mapped into a far richer space — without ever building that space., receptive fieldreceptive field, n.: The patch of the original input, such as a region of an image, that a particular neuron inside a network can see and respond to., stridestride, n.: How far a scanning window jumps each time it moves across an image. Jumping two pixels instead of one halves the size of the result., weight sharingweight sharing, n.: Reusing the same learned numbers in many places in a network, for example one edge detector applied at every spot in an image, so the network needs fewer parameters and learns from less data., depthwise separable convolutiondepthwise separable convolution, n.: A cheaper way to do the main operation of image networks: first scan each channel of the input separately, then mix the channels together with a tiny one-by-one step., dilated convolutiondilated convolution, n.: A version of the scanning-window operation that spreads its weights apart with gaps, so it can take in a wider area of an image or sound at no extra cost., transposed convolutiontransposed convolution, n.: An operation that roughly reverses an image network's shrinking steps, enlarging a small grid of features into a bigger one, as when a model paints a full-size image.