profile
viewpoint
If you are wondering where the data of this site comes from, please visit https://api.github.com/users/gtoubassi/events. GitMemory does not store any data, but only uses NGINX to cache data for a period of time. The idea behind GitMemory is simply to give users a better reading experience.

gtoubassi/femtozip 132

FemtoZip is a "shared dictionary" compression library optimized for small documents that may not compress well with traditional tools such as gzip

gtoubassi/dqn-atari 90

A TensorFlow based implementation of the DeepMind Atari playing "Deep Q Learning" agent that works reasonably well

gtoubassi/mnist-gan 25

A Generative Adversarial Network (GAN) for generating mnist digits

gtoubassi/mnist-vae 24

Semi-supervised learning with mnist using variational autoencoders. An unsupervised representation is learned which allows for superior classification results with limited labels.

gtoubassi/ProductClassification 15

A playground for classifying products based on image and text features using deep learning.

gtoubassi/mikado-universe 5

A javascript simulation of a "mikado universe" as described at http://bit.ly/ji8ai3. The toy universe shows how the macroscopic effect of gravity could be an "emergent" force resulting from the increase of entropy.

gtoubassi/simple-autodiff 3

A simple (but inefficient) auto diff algorithm in python using a "define by run" methodology.

gtoubassi/FileBunker 2

FileBunker is a file backup application which uses one or more GMail accounts as a free, offsite backup repository.

gtoubassi/JavaLaunch 2

JavaLaunch is yet another win32 java launcher designed for max simplicity. Built during creation of https://github.com/gtoubassi/FileBunker

gtoubassi/simple-raycasting 2

A simple implementation of ray casting like those used in early "3D-ish" games like Wolfenstein 3D.

push eventgtoubassi/simple-vanity-url-shortener

gtoubassi

commit sha cbe51afe94840e34270253588263a40bc85da1c4

Upgrade to newer packages and fix broken sheets api by upgrading to the latest google-spreadsheet lib

view details

push time in a month

push eventgtoubassi/femtozip

gtoubassi

commit sha 9c6f6b30c377473642dab771377a064220f19931

Update README.md

view details

push time in 2 months

push eventgtoubassi/femtozip

gtoubassi

commit sha 4b846c666ed633b3d7199b6a701e4fe242684473

Update README.md

view details

push time in 2 months

push eventgtoubassi/femtozip

gtoubassi

commit sha b70fc5d51cb430c6ea98039a9073e6a6775d52c2

Update README.md

view details

push time in 2 months

GollumEvent
GollumEvent
GollumEvent

push eventgtoubassi/femtozip

gtoubassi

commit sha e3a958dc2be237bb06e87b0d5aefd614c858ede0

Update README.md

view details

push time in 2 months

push eventgtoubassi/femtozip

gtoubassi

commit sha 02b4d9ea8425fd22c887f168401b90c4cb6b3b34

Update README.md

view details

push time in 2 months

push eventgtoubassi/femtozip

gtoubassi

commit sha ebfc44fa7ce638518e4544c66d78501305d48fae

Update README.md

view details

push time in 2 months

GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent
GollumEvent

push eventgtoubassi/femtozip

gtoubassi

commit sha 1951aeba84534e8eb621582e8ab815e9399069d3

Remove SDCH given it is deprecated, and pretty up some links

view details

push time in 2 months

push eventgtoubassi/femtozip

gtoubassi

commit sha 0f3cc75e103dc1cfbfebc63051091416b3d0f8a7

When no model types are specified, just build the femtozip model and use all the sample data vs this too-clever-by-half bakeoff between fz and a pure huffman entropy encoding only approach. It improves the results of the fzdatagen compression by a few percent, so whaddya know.

view details

push time in 2 months

push eventgtoubassi/femtozip

gtoubassi

commit sha 693b9f32e4b5f61ac408470318c45ddaade1ae73

Score a potential subtring included in the dictionary based on what its entropy encoded length in bits is, not its actual length. In other words if we are compressing english language, which has lots of e's, t's, a's, then its less interesting to include 'eats' than it is to include 'zouk'.

view details

push time in 2 months

push eventgtoubassi/femtozip

gtoubassi

commit sha 81b4bae0a4ae4c1b6be1028c14e785bec9d1cb8c

Don't pack 2x and truncate. This was an attempt to make sure to satisfy the desired dict length, but was actually counter productive because if you allow packing 200% the size you may end up doing worse on the first 100%.

view details

gtoubassi

commit sha 79388d6f8b95785b7cf6706a64efff5f83d273ac

Use a float score for value of a substring as the int score was a holdover from the java implementation which used a radix sort for performance. I was too lazy to port it to C++ which was prescient as it turned out to be a premature optimization. For some use cases I found that losing precision on the score mattered.

view details

push time in 2 months