Torch for R + luz
Torch for R… wow! 🔥🔥🔥 I was recently having a discussion with a coworker about the benefits of Torch, especially the power of training one global model capable of hierarchical projections (awesome for time series) and predicting multiple group-specific regressions. I went down a Googling rabbit hole last weekend and came across some amazing articles by Sigrid Keydana (see links below) introducing torch to the R community and also recently releasing luz, a high-level R interface to Torch (“luz is to torch what Keras is to TensorFlow”).
RStudio’s MLVerse team is doing really exciting things for the R machine learning and AI community. With torch, I no longer need to launch a conda environment for complex NNs (although having Python on your system is always handy 😅). And even better, “torch for R is built directly on top of libtorch, a C++ library that provides the tensor-computation and automatic-differentiation capabilities essential to building neural networks.” If you’re looking for fast NNs and deep learning solutions within the #rstats framework, give these packages a try. Happy Friday and happy learning! 🤓📚
Sources
- Torch for R: https://torch.mlverse.org/
- Intro to Torch: Torch for R
- Intro to Luz: Que haja luz: More light for torch!
- Intro to TabNet: TabNet for Tabular Data with Torch
- Intro to Time Series with Torch: Introductory Time-Series Forecasting with Torch
- Sigrid Keydana: Sigrid on LinkedIn