`trainnet` — Train a supported dlnetwork with a built-in loss.

trainnet(X,Y,net,lossFcn,options) trains a supported dlnetwork compatibility object over in-memory numeric or logical predictor data.

Syntax

net = trainnet(X, Y, net, lossFcn, options)
[net, info] = trainnet(X, Y, net, lossFcn, options)

How `trainnet` works

Supports the five-argument form with 2-D predictor matrices where rows are observations and columns are features.
net may be a dlnetwork, trained network compatibility object, simple layerGraph, or supported layer collection.
lossFcn supports "mse"/"mean-squared-error" for regression and "crossentropy" for classification.
For regression, Y must be a numeric or logical matrix with one row per observation and the same output width as the network.
For classification, Y may be numeric integer labels, string labels, categorical labels, or a one-hot/probability matrix. The returned dlnetwork stores sorted class labels in Classes when labels are provided.
trainingOptions metadata controls sgdm and adam training, including MaxEpochs, MiniBatchSize, InitialLearnRate, Momentum, Adam decay factors, Epsilon, deterministic Shuffle handling, and CPU/auto ExecutionEnvironment values.
The returned object is a dlnetwork compatibility object with updated layer weights, Learnables, TrainingInfo, TrainingLoss, and TrainedWith metadata.
Custom loss function handles are not executed by trainnet; use explicit dlfeval/dlgradient loops for custom host dlarray autodiff. Datastores, image/sequence/recurrent/convolutional/normalization/custom layers, branching DAG execution, validation data, callbacks, plots, and GPU training remain outside this slice and raise explicit compatibility errors or are stored as inert options.

GPU memory and residency

trainnet currently gathers gpuArray inputs to host before training when the execution environment is auto or cpu. Explicit ExecutionEnvironment values of gpu, multi-gpu, or parallel raise a compatibility error because native GPU training kernels, provider-resident Deep Learning layers, and GPU-backed autodiff remain future acceleration work.

Example

Train a dlnetwork with crossentropy

net = dlnetwork(layers);
opts = trainingOptions('adam','MaxEpochs',120,'MiniBatchSize',4,'InitialLearnRate',0.05,'Shuffle','never');
net = trainnet(X, labels, net, 'crossentropy', opts)

Expected output:

net is a trained dlnetwork compatibility object usable with forward or predict.

Using `trainnet` with coding agents

Open a RunMat example with live inputs, then ask the agent to explain how trainnet changes the result.

Run a small trainnet example, explain the result, then change one input and compare the output.

adamupdate · analyzeNetwork · bilstmLayer · classificationLayer · combvec · convolution1dLayer · crossentropy · dlarray · dlfeval · dlgradient · dlnetwork · dlupdate · eluLayer · exportONNXNetwork · featureInputLayer · forward · fullyConnectedLayer · globalAveragePooling1dLayer · layerGraph · layerNormalizationLayer · lstmLayer · padsequences · regressionLayer · reluLayer · sequenceInputLayer · softmaxLayer · trainingOptions · trainNetwork

About RunMat

RunMat is an open-source runtime that executes MATLAB-syntax code blazing on any GPU. It is licensed under the Apache 2.0 license.

RunMat automatically optimizes your math for GPU execution on Apple, Nvidia, and AMD hardware. No code changes needed. Simulations that took hours now take minutes.
Start running code in seconds. RunMat runs in the browser, on the desktop, or from the CLI. No license server, no IT ticket.

Getting started · Benchmarks · Pricing

trainnet — Train a supported dlnetwork with a built-in loss.

Syntax

How trainnet works