0 Persons are tagged with #compression

#compression

One person like that
3 Likes
2 Comments
waynerad@diasp.org
Wayne Radinsky

Using compression algorithms to do text classification, competitive with deep neural networks. Neural networks have proven so effective at so many things, sometimes it's interesting to see there are non-neural-network solutions that can compete with them or even outcompete them.

The theory here is that text compression algorithms work by using information theory to reduce redundancy within a text. When combining texts, it can be used to approximate the information distance between two texts. It should be noted that it can only be approximated because no compression algorithm can be proven to be the maximum compressor.

To make an actual classification system, they combined the compression algorithm with something called kNN. "kNN" stands for k-nearest-neighbors. The idea is you pick some number "k" which is the number of clusters that you want. The algorithm then figures out for itself how to group the items to be classified into k clusters. To do this, it needs a "distance" measure.

Here they tried bz2, lzma, zstd, and gzip, and found gzip did the best.

They then compared gzip with the following neural network-based text classification systems: TFIDF+LR, LSTM, Bi-LSTM+Attn, HAN, charCNN, textCNN, RCNN, VDCNN, fastText, BERT, W2V, SentBERT, and TextLength. They tested them on the following datasets: AGNews (academic news), DBpedia (extracted from Wikipedia), YahooAnswers (from Yahoo obviously), 20News (an old news dataset from 1995), Ohsumed (news from medical journals between 1987 and 1991), R8 and R52 (two datasets of news from Reuters), KirundiNews and KinyarwandaNews (two datasets of news in low-resource African languages), SwahiliNews (news in Swahili, a language from east Africa), DengueFilipino (news in Filipino), and SogouNews (news in Chinese, written in pinyin).

The only neural network that consistently outperformed their gzip+kNN system was BERT.

Oh, but there's a catch. BERT only beat the gzip+kNN system when classifying data similar to what it was trained on. When classifying text significantly different from what it was trained on, "out-of-distribution" data in the parlance of statisticians, BERT actually did worse, and the gzip+kNN system beat everything. In addition, the gzip+kNN system requires less computing power.

"Low-resource" text classification: A parameter-free classification method with compressors

#solidstatelife #ai #informationtheory #compression #gzip #knn

“Low-Resource” Text Classification: A Parameter-Free Classification Method with Compressors

Zhiying Jiang, Matthew Yang, Mikhail Tsirlin, Raphael Tang, Yiqin Dai, Jimmy Lin. Findings of the Association for Computational Linguistics: ACL 2023. 2023.

5 Likes
5 Comments
4 Shares
petapixel@xn--y9azesw6bu.xn--y9a3aq
petapixel (unofficial)

Nikon Z9 to Shoot 8K at 60FPS Thanks to New Compressed RAW Codec

image

A high-efficiency RAW recording codec developed by intoPIX has been successfully integrated into the Nikon Z9. The company says that its TicoRAW technology allows for low processing and fast transfer speeds of up to 8K at 60 frames per second in RAW.

Nikon formed a strategic partnership with intoPIX in order to leverage the company's RAW compression technology, TicoRAW. The company claims this technology delivers the most efficient RAW files that do not compromise image integrity at all thanks to "innovative processing and coding." intoPIX says the full power of an image sensor is preserved while TicoRAW is able to reduce bandwidth, storage needs, and transfer times.

It is thanks to TicoRAW that the Nikon Z9 will be able to capture 8K at up to 60 frames per second, which will come to the camera in a 2022 firmware update as noted at the time of announcement. The reason this capability wasn't available at launch might have been because intoPIX had not yet finalized the implementation of the RAW codec into the camera by the October announcement, but today intoPIX has confirmed that it has successfully done so.

ticoraw high efficiency codec in nikon z9 menu

"TicoRAW is the world’s first RAW codec, handling both RAW pictures and RAW videos, that can offer compression efficiency with such low complexity on RAW Bayer and other Color Filter Arrays (CFA) patterns," intoPIX says.

intopix ticoraw nikon z9 graphic

TicoRAW is a patented technology that intoPIX claims allows for the capture of high image quality as well as the capability to manage very high resolutions, high frame rates, and high dynamic range workflows.

"TicoRAW is the world’s first RAW codec that can offer compression efficiency with such low complexity. It also has the advantage of retaining this very fast format for editing, regardless of resolution and/or frame rates used," the company adds.

intoPIX says TicoRAW is capable of supporting a huge number of resolutions, from 1080p Full HD through 20K RAW.

intoPIX doesn't just say the technology is limited to the Z9, though that is where the codec will make its first appearance. The company broadly states that it has successfully integrated the TicoRAW technology into "the new generation of Nikon cameras" which happens to start with the Z9, though the company appears to leave the door open to see its technology leveraged on other Nikon cameras.

#equipment #news #technology #8kp60 #8kraw #codec #compression #intopix #nikonz9 #nikonz #nikonzmount #raw #rawcodec #rawvideo #ticoraw #z9

Nikon Z9 to Shoot 8K at 60FPS Thanks to New Compressed RAW Codec

This is how Nikon is going to give the Z9 the ability to shoot 8K at 60FPS RAW video.

1 Shares