Author: T. Moudiki

Feed: R-bloggers. Author: T. Moudiki. This post is about LSBoost, an Explainable ‘AI’ algorithm which uses Gradient Boosted randomized networks for pattern recognition. As we’ve discussed it last week LSBoost is a cousin of GFAGBM’s LS_Boost. In LSBoost, more specifically, the so called weak learners from LS_Boost are based on randomized neural networks’ components and variants of Least Squares regression models.I’ve already presented some promising examples of use of LSBoost based on Ridge Regression weak learners. In mlsauce’s version 0.7.1, the Lasso can also be used as an alternative ingredient to the weak learners. Here is a comparison of the ... Read More

Feed: R-bloggers. Author: T. Moudiki. Disclaimer: I have no affiliation with The Next Web (cf. online article)A few weeks ago I read this interesting and accessible article about explainable AI, discussing more specifically self-explainable AI issues. I’m not sure – anymore – if there’s a mandatory need for AI models that explain themselves, as there are model-agnostic tools such as the teller – among many others – for helping them in doing just that.With that being said, the new LSBoost algorithm implemented in mlsauce does, explain itself. LSBoost is a cousin of the LS_Boost algorithm introduced inGREEDY FUNCTION APPROXIMATION: A ... Read More

Feed: R-bloggers. Author: T. Moudiki. nnetsauce is a general purpose tool for Statistical/Machine Learning, in which pattern recognition is achieved by using quasi-randomized networks. A new version, 0.5.0, is out on Pypi and for R:Install by using pip (stable version):pip install nnetsauce --upgrade Install from Github (development version):pip install git+https://github.com/thierrymoudiki/nnetsauce.git --upgrade Install from Github, in R console:library(devtools) devtools::install_github("thierrymoudiki/nnetsauce/R-package") library(nnetsauce) This could be the occasion for you to re-read all the previous posts about nnetsauce, or to play with various examples in Python or R. Here are a few other ways to interact with the nnetsauce:1) FormsIf you’re not comfortable with ... Read More

Feed: R-bloggers. Author: T. Moudiki. # tips' classes = response variable y_int = np.asarray([0, 0, 2, 2, 2, 3, 0, 2, 0, 2, 0, 3, 0, 1, 2, 2, 0, 2, 2, 2, 3, 1, 1, 3, 2, 1, 0, 0, 3, 1, 0, 1, 1, 1, 2, 2, 0, 2, 1, 3, 1, 1, 2, 0, 3, 1, 3, 3, 1, 1, 1, 1, 3, 0, 3, 2, 1, 0, 0, 3, 2, 0, 0, 2, 1, 2, 1, 0, 1, 1, 0, 1, 2, 3, 1, 0, 2, 2, 1, 1, 1, 2, 0, 3, 1, 3, 0, ... Read More

Feed: R-bloggers. Author: T. Moudiki. A few weeks ago, I introduced a target-based categorical encoder for Statistical/Machine Learning based on correlations + Cholesky decomposition. That is, a way to convert explanatory variables such as the x below, to numerical variables which can be digested by ML models. # Have: x c("apple", "tomato", "banana", "apple", "pineapple", "bic mac", "banana", "bic mac", "quinoa sans gluten", "pineapple", "avocado", "avocado", "avocado", "avocado!", ...) # Need: new_x c(0, 1, 2, 0, 3, 4, 2, ...) This week, I use the tips dataset (available here). Imagine that you work in a restaurant, and also have ... Read More

Feed: R-bloggers. Author: T. Moudiki. Last week on this blog, I presented AdaOpt for R, applied to iris dataset classification. And the week before that, I introduced AdaOpt for Python. AdaOpt is a novel probabilistic classifier, based on a mix of multivariable optimization and a nearest neighbors algorithm. More details about the algorithm can be found in this (short) paper. This week, we are going to train AdaOpt on the popular MNIST handwritten digits dataset without preprocessing, a.k.a neither convolution nor pooling. Install mlsauce’s AdaOpt from the command line (for R, cf. below): !pip install git+https://github.com/thierrymoudiki/mlsauce.git --upgrade Import the ... Read More

Feed: R-bloggers. Author: T. Moudiki. Last week on this blog, I presented AdaOpt for Python on a handwritten digits classification task. AdaOpt is a novel probabilistic classifier, based on a mix of multivariable optimization and a nearest neighbors algorithm. It’s still very new and only time will allow to fully appreciate all of its features. The tool is fast due to Cython, and to the ubiquitous (and mighty) numpy, which both help in bringing C/C++ -like performances to Python. There are also a few tricks available in AdaOpt, that allow to make it faster to train on bigger datasets ... Read More

Feed: R-bloggers. Author: T. Moudiki. This post is about using custom error measures in crossval, a tool offering generic functions for the cross-validation of Statistical/Machine Learning models. More information about cross-validation of regression models using crossval can be found in this post, or this other one. The default error measure for regression in crossval is Root Mean Squared Error (RMSE). Here, I’ll show you how to obtain two other error measures: Mean Absolute Percentage Error (MAPE) Mean Absolute Error (MAE) The same principles can be extended to any other error measure of your choice. Installation of crossval From Github, ... Read More

Feed: R-bloggers. Author: T. Moudiki. Sometimes in Statistical/Machine Learning problems, we encounter categorical explanatory variables with high cardinality. Let’s say for example that we want to determine if a diet is good or bad, based on what a person eats. In trying to answer this question, we’d construct a response variable containing a sequence of characters good or bad, one for each person; and an explanatory variable for the model would be: x = c("apple", "tomato", "banana", "apple", "pineapple", "bic mac", "banana", "bic mac", "quinoa sans gluten", "pineapple", "avocado", "avocado", "avocado", "avocado!", ...) Some Statistical/Machine learning models only accept numerical ... Read More

Feed: R-bloggers. Author: T. Moudiki. As seen last week in a post on grid search cross-validation, crossval contains generic functions for statistical/machine learning cross-validation in R. A 4-fold cross-validation procedure is presented below: In this post, I present some examples of use of crossval on a linear model, and on the popular xgboost and randomForest models. The error measure used is Root Mean Squared Error (RMSE), and is currently the only choice implemented. Installation From Github, in R console: devtools::install_github("thierrymoudiki/crossval") Demo We use a simulated dataset for this demo, containing 100 examples, and 5 explanatory variables: # dataset creation set.seed(123) ... Read More