PyLaia is the successor of Laia, ported to PyTorch and including several improvements over the Lua-based version.

Laia allows you to train convolutional and recurrent neural network models using CTC for recognition applications like handwritten text recognition (HTR). It was built using the popular Torch framework, which allows you to use seamlessly both the CPU and GPUs.

This is a fast and memory efficient implementation of the Principal Components Analysis algorithm.

  • Why fast? First, it uses BLAS + LAPACK to perform matrix operations efficiently. Secondly, you can compute the principal components or reduce the dimensionality of your data with just a single pass over it. Plus, you can split your data into multiple chunks and process it in parallel.
  • Why memory efficient? You don’t need to load the whole dataset into main memory not even a single time. Plus, you can use simple or double precision.

Just another simple compressor/decompressor supporting Huffman/LZ77/LZ78/LZW algorithms.

I wrote this program only for educational purposes, during the Coding Theory course at the Universitat Politècnica de València. If you want to use a real compressor, use gzip or bzip2 instead.