Choose an AI Model

This page provides guidance on choosing an AI model from the options across many MATLAB^® toolboxes.

Note

For information about reinforcement learning models, see What Is Reinforcement Learning? (Reinforcement Learning Toolbox).

Data Considerations

Consult this flow chart to determine a suggested AI model.

Start of a flowchart for determining a suggest AI model. For image data tasks, click link 1. For text data tasks, click link 2. For tabular and small time-series data tasks, click link 3. For time-series and large feature data tasks, click link 4.

Image Tasks

Flowchart for choosing an AI model for image data tasks. For predicting, click link 1. For generating images, click link 2. For classifying images, click link 3. For segmenting images, click link 4. For detecting objects, click link 5.

End point	Suggested Models
1. Predict values	Convolutional networks: Train Convolutional Neural Network for Regression Convert Classification Network into Regression Network Build Image-to-Image Regression Network Using Deep Network Designer Maritime Clutter Suppression with Neural Networks (Radar Toolbox)
2. Generate images	Diffusion model: Generate Images Using Diffusion Adversarial networks: Train Generative Adversarial Network (GAN) Train Conditional Generative Adversarial Network (CGAN)
3. Classify entire images	Pretrained convolutional networks: Pretrained Deep Neural Networks Classify Image Using GoogLeNet Convolutional networks: Create Simple Deep Learning Neural Network for Classification Activity Recognition from Video and Optical Flow Data Using Deep Learning Gesture Recognition using Videos and Deep Learning Hand Gesture Classification Using Radar Signals and Deep Learning
4. Classify pixels (semantic segmentation)	Convolutional networks: Semantic Segmentation of Multispectral Images Using Deep Learning (Image Processing Toolbox) 3-D Brain Tumor Segmentation Using Deep Learning Brain MRI Segmentation Using Pretrained 3-D U-Net Network Aerial Lidar Semantic Segmentation Using PointNet++ Deep Learning (Lidar Toolbox) Aerial Lidar Semantic Segmentation Using RandLANet Deep Learning (Lidar Toolbox) Lidar Point Cloud Semantic Segmentation Using SqueezeSegV2 Deep Learning Network (Lidar Toolbox) Spectrum Sensing with Deep Learning to Identify 5G, LTE, and WLAN Signals (Communications Toolbox)
5. Detect objects and separate instances of objects (instance segmentation)	Overview: Get Started with Object Detection Using Deep Learning (Computer Vision Toolbox) YOLO models: Multiclass Object Detection Using YOLO v2 Deep Learning (Computer Vision Toolbox) Object Detection Using YOLO v3 Deep Learning (Computer Vision Toolbox) Object Detection Using YOLO v4 Deep Learning (Computer Vision Toolbox) Object Detection in Large Satellite Imagery Using Deep Learning (Computer Vision Toolbox) Lidar Object Detection Using Complex-YOLO v4 Network Regions with convolutional neural networks (R-CNNs): Getting Started with R-CNN, Fast R-CNN, and Faster R-CNN (Computer Vision Toolbox) Train Object Detector Using R-CNN Deep Learning (Computer Vision Toolbox) Perform Instance Segmentation Using Mask R-CNN (Computer Vision Toolbox) Convolutional network: Lidar 3-D Object Detection Using PointPillars Deep Learning (Lidar Toolbox)

Text Tasks

Flowchart for choosing an AI model for text data tasks. For generating text, click link 1. For parsing grammar, click link 2. For classifying or summarizing text, click link 3. For analyzing sentiment, click link 4. For modeling topics, click link 5. For recognizing named entities, click link 6. For extracting keywords, click link 7. For retrieving information, click link 8.

End point	Suggested Models
1. Generate text	LSTM network: Word-by-Word Text Generation Using Deep Learning (Text Analytics Toolbox) Autoencoder: Generate Text Using Autoencoders (Text Analytics Toolbox) Transformer models: Summarize Text Using Transformers (GitHub) Large Language Models (LLMs) with MATLAB (GitHub)
2. Parse grammar	Transformer model: Analyze Sentence Structure Using Grammatical Dependency Parsing (Text Analytics Toolbox)
3. Classify or summarize text	Transformer models: Train BERT Document Classifier (Text Analytics Toolbox) Fine-Tune Pretrained BERT Model (GitHub) Bag-of-words model: Create Simple Text Model for Classification (Text Analytics Toolbox) Machine learning models: Classify Documents Using Document Embeddings (Text Analytics Toolbox) LSTM network: Classify Text Data Using Deep Learning (Text Analytics Toolbox) Summarization algorithms: `extractSummary` (Text Analytics Toolbox)
4. Analyze sentiment	Built-in sentiment analysis functions (VADER algorithm): Sentiment Analysis in MATLAB (Text Analytics Toolbox) Machine learning models: Train a Sentiment Classifier (Text Analytics Toolbox) Transformer models: Analyze Sentiment Using FinBERT (GitHub)
5. Model topics	Latent Dirichlet allocation (LDA): Analyze Text Data Using Topic Models (Text Analytics Toolbox) Latent semantic analysis: `fitlsa` (Text Analytics Toolbox)
6. Recognize named entities	Hidden Markov model (HMM): Train Custom Named Entity Recognition Model (Text Analytics Toolbox)
7. Extract keywords	RAKE algorithm: Extract Keywords from Text Data Using RAKE (Text Analytics Toolbox) TextRank algorithm: Extract Keywords from Text Data Using TextRank (Text Analytics Toolbox)
8. Retrieve information	Best-matching similarity: Information Retrieval with Work Orders Data (Text Analytics Toolbox) `bm25Similarity` (Text Analytics Toolbox)

Tabular Data and Small Time-Series Tasks

Flowchart for choosing an AI model for tabular and small time-series data tasks. For unsupervised methods, click link 1. For semi-supervised methods, click link 2. For classifying data, click link 3. For predicting values, click link 4. For forecasting future values, click link 5.

End point	Suggested Models
1. Use unsupervised methods	Cluster Analysis: Choose Cluster Analysis Method (Statistics and Machine Learning Toolbox) Labeling data: Choose an App to Label Ground Truth Data (Computer Vision Toolbox) Dimensionality Reduction and feature transformation: Reduce Dimensionality (Statistics and Machine Learning Toolbox) Visualize High-Dimensional Data Using t-SNE (Statistics and Machine Learning Toolbox) Nonnegative Matrix Factorization (Statistics and Machine Learning Toolbox) `incrementalPCA` (Statistics and Machine Learning Toolbox) Anomaly detection models: Unsupervised Anomaly Detection (Statistics and Machine Learning Toolbox) Incremental Anomaly Detection Overview (Statistics and Machine Learning Toolbox) Configure Model for Incremental Anomaly Detection (Statistics and Machine Learning Toolbox) Anomaly Detection with Isolation Forest (Statistics and Machine Learning Toolbox)
2. Use semi-supervised methods	Classification models: Label Data Using Semi-Supervised Learning Techniques (Statistics and Machine Learning Toolbox)
3. Classify data	Classification Learner app: Classification Learner (Statistics and Machine Learning Toolbox) Choose Classifier Options (Statistics and Machine Learning Toolbox) Train Classification Models in Classification Learner App (Statistics and Machine Learning Toolbox) Command-line classification models: Choose an Algorithm (Statistics and Machine Learning Toolbox) Characteristics of Classification Algorithms (Statistics and Machine Learning Toolbox) Train Neural Network with Tabular Data Incremental learning models: Implement Incremental Learning for Classification Using Succinct Workflow (Statistics and Machine Learning Toolbox) Initialize Incremental Learning Model from Logistic Regression Model Trained in Classification Learner (Statistics and Machine Learning Toolbox) Anomaly detection models: Model-Specific Anomaly Detection (Statistics and Machine Learning Toolbox)
4. Predict values (regression)	Regression Learner app: Regression Learner (Statistics and Machine Learning Toolbox) Choose Regression Model Options (Statistics and Machine Learning Toolbox) Train Regression Models in Regression Learner App (Statistics and Machine Learning Toolbox) Command-line regression models: Choose an Algorithm (Statistics and Machine Learning Toolbox) Incremental learning models: Implement Incremental Learning for Regression Using Succinct Workflow (Statistics and Machine Learning Toolbox) Initialize Incremental Learning Model from SVM Regression Model Trained in Regression Learner (Statistics and Machine Learning Toolbox)
5. Forecast future values	Direct forecasting: Perform Time Series Direct Forecasting with directforecaster (Statistics and Machine Learning Toolbox) Regression trees: Manually Perform Time Series Forecasting Using Ensembles of Boosted Regression Trees (Statistics and Machine Learning Toolbox)

Time-Series and Large Feature Data Tasks

Flowchart for choosing an AI model for time-series and large feature data tasks. For generating data, click link 1. For predicting single values, click link 2. For predicting sequences, click link 3. For forecasting future values, click link 4. For classifying entire samples, click link 5. For classifying time steps, click link 6.

End point	Suggested Models
1. Generate data	Adversarial networks: Generate Synthetic Signals Using Conditional GAN Train Generative Adversarial Network (GAN) for Sound Synthesis
2. Predict single values (sequence-to-one regression)	LSTM networks: Sequence-to-One Regression Using Deep Learning Long Short-Term Memory Neural Networks Convolutional networks: Pretrained Neural Networks for Audio Applications Multilayer perceptron networks: Neural Network for Digital Predistortion Design-Offline Training (Communications Toolbox)
3. Predict sequences (sequence-to-sequence regression)	LSTM networks: Sequence-to-Sequence Regression Using Deep Learning Long Short-Term Memory Neural Networks Physical System Modeling Using LSTM Network in Simulink Neural ODE network: Dynamical System Modeling Using Neural ODE Transformer network: `speech2text` (Audio Toolbox)
4. Forecast future values	LSTM networks: Time Series Forecasting Using Deep Learning Build Time Series Forecasting Network Using Deep Network Designer Long Short-Term Memory Neural Networks
5. Classify entire samples	LSTM networks: Sequence Classification Using Deep Learning Spoken Digit Recognition with Wavelet Scattering and Deep Learning (Audio Toolbox) Convolutional networks: Sequence Classification Using 1-D Convolutions Spoken Digit Recognition with Wavelet Scattering and Deep Learning (Audio Toolbox) Modulation Classification with Deep Learning (Communications Toolbox) Convolutional and LSTM network: Sequence Classification Using CNN-LSTM Network Multilayer perceptron networks: Train Speech Command Recognition Model Using Deep Learning Train Neural Network with Tabular Data Autoencoder network: OFDM Autoencoder for Wireless Communications (Communications Toolbox)
6. Classify time steps (sequence-to-sequence classification)	LSTM networks: Sequence-to-Sequence Classification Using Deep Learning Long Short-Term Memory Neural Networks Autoencoders: Time Series Anomaly Detection Using Deep Learning Detect Anomalies in Signals Using deepSignalAnomalyDetector Audio-Based Anomaly Detection for Machine Health Monitoring Detect Anomalies in Industrial Machinery Using Three-Axis Vibration Data