3 ms·

>machine learning development has just stopped dead From https://reference.wolfram.com/language/guide/SummaryOfNewFeaturesIn131 https://reference.wolfram.com/l

>machine learning development has just stopped dead From https://reference.wolfram.com/language/guide/SummaryOfNewFeaturesIn131 https://reference.wolfram.com/language/guide/SummaryOfNewFea... New in machine learning in 13.1 ... ContentDetectorFunction (updated) — support for Information DimensionReducerFunction(updated) ▪ FeatureExtractorFunction(updated) FindClusters (updated) — support for fixed number of clusters and UpTo ClusterClassify(updated) ▪ ClusteringComponents(updated) Interpretable Machine Learning FeatureValueImpactPlot — plots the impact of a feature on a model result FeatureImpactPlot — plots the impact of each feature together CumulativeFeatureImpactPlot — plots the cumulative impact of each feature FeatureValueDependencyPlot — plots the result dependency on a feature value Network Layers » ReplicateLayer (updated) — support for integer arrays RandomArrayLayer (updated) — support for more statistical distribution AttentionLayer (updated) — support for dropout and local masking ElementwiseLayer (updated) — new activations "Mish" and "GELU" ThreadingLayer(updated) ▪ FunctionLayer(updated) Network Training » NetTrain (updated), LossFunction (updated) — support for multi-output and nonscalar loss Encoders & Decoders » "Image" (updated) — resampling and padding support "Class" (updated) — support for top-k and top-p sampling (nucleus sampling) Formats "ONNX" (updated) — export support for net operators

n-e-w 4y ago
Three of their best ML engineers /devs left in the last 12 months or so. The backend for anything deep-learning is MXNet...which is essentially a non-starter these days vs PyTorch or TensorFlow. All these functions you cite are very low level updates not real feature updates. I would say in 2018-2019 the ML / Deep learning stack was actually very impressive compared with what was available in FOSS. Now -- it's languishing. Badly.

n-e-w 4y ago

Three of their best ML engineers /devs left in the last 12 months or so. The backend for anything deep-learning is MXNet...which is essentially a non-starter these days vs PyTorch or TensorFlow. All these functions you cite are very low level updates not real feature updates. I would say in 2018-2019 the ML / Deep learning stack was actually very impressive compared with what was available in FOSS. Now -- it's languishing. Badly.