A library for machine learning and data analysis in R
R Machine Learning (Rml) is a machine learning library for the R programming language that I have begun to refactor from the work I did during my Master’s Thesis. Rml is not currently available to the public, but it will be once I have things cleaned up and documented. Stay tuned!!
Rml will include a number of machine learning algorithms with a focus on neural networks, including:
- Feedforward networks
- Elman style recurrent networks
- Echo State Networks
- Various optimization algorithms, such as Scaled Conjugate Gradients, RProp, Stochastic Gradient Descent and Alopex
- Principal Components Analysis
- Linear and Quadratic Discriminant Analysis
- Linear Logistic Regression
- Autoregressive Models
A feedforward network trained with ALOPEX learning to fit a noisy standing sine wave.
A feedforward network trained with ALOPEX learning to fit the exclusive or (XOR) function.
A two-layer feedforward network trained using Scaled Conjugate Gradients (SCG) to fit a noisy sinc function.
An Echo State Network (ESN) with 500 hidden units solving the temporal XOR problem with a shift of 6. This problem requires both a non-linear transfer function as well as state/memory.