cyrilou242 / ftcc

Fast Text Classification with Compressors dictionary
MIT License
146 stars 10 forks source link

FTCC with C# #5

Open techjb opened 1 year ago

techjb commented 1 year ago

I have created a C# version for this repository: FTCC C#

Features:

cyrilou242 commented 1 year ago

Very nice! Are you able to reproduce the accuracy results with a compression level of 9?

Also I'm curious about the speed you achieve, python is very slow at string concatenation, this is taking a significant amount of time in the training phase. Not sure for C# (and I don't know how the JIT works in C#)

Feel free to add PR to this project to add to the README something like:

## Other implementations
- [C#](link-to-your-repo)
techjb commented 1 year ago

I have found hardly any variation in the results with compression level 9. However, I have noticed a slight decrease in speed when predicting. For the creation of the dictionaries, I used the ZstdNet library, which takes a list of byte arrays as input, so it is not necessary to concatenate strings. I think that overall the performance in terms of speed is very good.