Machine Learning

ml.svm ml.svm implements a support vector machine for classifying feature vectors.
ml.softmax ml.softmax is a softmax classifier that trains on labeled feature vectors added with add and classifies an input feature vector using map.
ml.randforest ml.randforest implements Random Forest classification, training a model from labeled feature vectors and mapping input feature vectors to output values.
fluid.skmeans fluid.skmeans implements the K-means clustering algorithm with cosine similarity.
fluid.kmeans The fluid.kmeans object implements the K-means clustering algorithm, used to group data points from a fluid.dataset into a specified number of clusters.
ml.dtw ml.dtw uses Dynamic Time Warping to train a model on labeled time-series feature vectors and map input vectors to classification output.
ml.adaboost ml.adaboost performs classification using Adaptive Boosting, training a model from labeled feature vectors added with add and mapping input feature vectors to output values with map.
ml.mindist ml.mindist implements minimum-distance classification, training a model from labeled feature vectors added as lists whose first value is the class ID.
fluid.labelset The fluid.labelset object is a container that associates string identifiers with corresponding labels, primarily used in conjunction with fluid.dataset~ for machine learning tasks.
fluid.nmfmorph~ The fluid.nmfmorph~ object performs cross-synthesis between two audio sources using Nonnegative Matrix Factorization (NMF) and Optimal Transport (OT).
ml.knn ml.knn implements K-nearest-neighbour classification, mapping an input feature vector to output value(s).
pix_opencv_haarcascade pix_opencv_haarcascade detects objects in image frames using a trained cascade classifier, reporting the number of detected objects and, for each detection, its center coordinates and radius.
fluid.umap The fluid.umap object performs Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction on a fluid.dataset.
fluid.plotter The fluid.plotter object provides a native graphical user interface for plotting and querying two-dimensional datasets.
pix_opencv_trackKnn pix_opencv_trackKnn uses the KNN algorithm to keep track of detected features supplied as cvblob lists, and can be used with other tracking techniques.
ml.hmmc ml.hmmc implements a continuous Hidden Markov Model for classifying input feature-vector time series.
onnx onnx is a lightweight Pure Data external that lets you run ONNX models directly inside Pd.
fluid.knnregressor The fluid.knnregressor object performs K-Nearest Neighbours regression, enabling the prediction of outputs by finding the weighted average of neighboring data points within datasets.
fluid.mds The fluid.mds object performs Multidimensional Scaling (MDS) for dimensionality reduction on a fluid.dataset.
fluid.mlpregressor The fluid.mlpregressor object performs regression using a Multi-Layer Perceptron (MLP) neural network.
fluid.nmffilter~ The fluid.nmffilter~ object decomposes and resynthesizes an audio signal by filtering it against spectral templates, or "bases," typically generated by fluid.bufnmf~.
fluid.bufnmfcross fluid.bufnmfcross reconstructs a target sound by using components learned from a source sound through Non-negative Matrix Factorization (NMF).
choice choice stores a list of vectors and searches them for the best match to an incoming list, outputting the matching vector’s index.
ml.gmm ml.gmm is a Gaussian Mixture Model classifier that trains on labeled feature vectors and maps input feature vectors to classification output.
fluid.bufnmfseed The fluid.bufnmfseed object computes initial bases and activations for Non-negative Matrix Factorization (NMF) from an input audio buffer.
ml.linreg ml.linreg implements linear regression, training a model from labeled feature vectors added with add and generating output values for an input feature vector with map.
ml.logreg ml.logreg implements logistic regression for classifying feature vectors.
py.o.train py.o.train trains ONNX models for timbre recognition in OpenScofo from MIR descriptors extracted from categorized training sounds.
fluid.nmfmatch~ The fluid.nmfmatch~ object matches an incoming audio signal against a set of spectral templates, typically derived using Non-negative Matrix Factorization (NMF).
pix_opencv_knear pix_opencv_knear performs OCR-like pattern recognition by comparing an incoming frame with a set of image patterns and outputting the comparison distance.
fluid.knnclassifier The fluid.knnclassifier object implements a K-Nearest Neighbors classifier, leveraging a fluid.kdtree for efficient data searching.
ml.dtree ml.dtree implements decision-tree classification, training a model from labeled feature vectors added with add and generating results for new vectors with map.
fluid.bufnmf The fluid.bufnmf object performs Non-Negative Matrix Factorisation (NMF) to decompose the spectrum of a sound into multiple components, represented by 'bases' (spectral profiles) and 'activations' (temporal envelopes).
fluid.standardize The fluid.standardize object rescales a FluidDataSet by transforming its data to have a mean of 0 and a standard deviation of 1 for each dimension.
fluid.bufflatten The fluid.bufflatten object flattens a multi-channel buffer into a single-channel buffer.
fluid.ls2text fluid.ls2text is a utility object designed to convert a fluid.labelset from a fluid.dataset into a text object.
fluid.robustscale The fluid.robustscale object applies robust scaling to fluid.dataset objects.
neuralnet~ neuralnet~ is the signal-rate version of neuralnet, providing a feed-forward artificial neural network for classification, regression, and binary logistic regression by loading models trained with neuralnet.
fluid.mlpclassifier The fluid.mlpclassifier object performs classification using a Multi-Layer Perceptron (MLP) neural network.
fluid.pca fluid.pca performs Principal Component Analysis (PCA) on a fluid.dataset to reduce dimensionality, such as transforming 13 MFCC values into 2.
ml.ann ml.ann implements an artificial neural network for classification or regression, trained from vectors supplied with add.
fluid.normalize The fluid.normalize object normalizes a fluid.dataset or a single data point.