Functionality for AI

Functions and objects that support deep learning workflows for wireless applications

With Communications Toolbox™ and Deep Learning Toolbox™, you can train wireless communications systems that contain a neural network.

Use AI functionality to analyze and optimize wireless channels for better signal processing, channel estimation, and resource management.
Training a wireless neural network requires Communications Toolbox functions and objects that support the dlarray (Deep Learning Toolbox) data objects. For the set of wireless communications domain-specific functions that support dlarray objects, see Wireless Communications (Deep Learning Toolbox).

Functions

`awgn`	Add white Gaussian noise to signal
`bit2int`	Convert bits to integers (Since R2021b)
`genqammod`	General quadrature amplitude modulation (QAM)
`ofdmmod`	Modulate signals using OFDM method
`ofdmdemod`	Demodulate signals using OFDM method
`ofdmChannelResponse`	Calculate OFDM channel response (Since R2023a)
`ofdmEqualize`	Equalize OFDM signals (Since R2022b)

`comm.ChannelFilter`	Filter signal using multipath gains at specified path delays
`comm.MIMOChannel`	Filter input signal through MIMO multipath fading channel

`dpdPreprocessor`	Digital predistortion (DPD) data input preprocessor for AI algorithms (Since R2024b)
`signalDatastore`	Datastore for collection of signals

Background Data Generation
Generate simulation input data in the background while running your simulation loop on parameterized conditions rather than generating the data in advance and recalling the stored data and run time. (Since R2026a)
Cosine Similarity As a Channel Estimate Quality Metric
Use the cosine similarity metric to compare two vectors. (Since R2024b)
Normalized Mean Squared Error as a Distance Measure
Use the normalized mean squared error (NMSE) as a loss function for training a neural network in a wireless communications application. (Since R2025a)