6 Tried-and-True Ways to Apply AI

The possibilities of AI are so vast, it can be easy to glide over the fact that multiple machine learning algorithms may be used in different parts of the system to build one “AI.” This slideshow presents six foundational tasks you can do with AI, along with examples and common algorithms. These tasks use techniques found in reinforcement learning, deep learning, and traditional machine learning and are the building blocks of many modern AI applications.

  • Enhance Images and Signals

    Applications: Improve image resolution, denoise signals in audio, create augmented images

    Input: Images and signal data

    Common algorithms: LSTM, CNNs, VDSR neural network

    Example: Perform domain translation between images. The Day-to-Dusk Image Translation example uses an unsupervised image-to-image translation network.

  • Identify Objects or Actions in Image, Video, and Signal Data

    Applications: Advanced driver assistance systems (ADAS) with object detection, semantic segmentation, robotics, computer vision perception for image recognition, activity detection, voice biometrics (voiceprint), keyword detection, smart devices

    Input: Images, videos, signals

    Common algorithms: CNNs with YOLO, clustering, Viola-Jones

    Example: Classify every pixel in a street-level image. The Semantic Segmentation Using Deep Learning example uses the Deeplab v3+ [1] CNN.

  • Move an Object Physically or in a Simulation

    Applications: Control systems, robotics in manufacturing, self-driving cars, drones, video games

    Input: Mathematical models, sensor data, videos, lidar data

    Common algorithms: Reinforcement learning, artificial neural networks (ANNs), CNNs, recurrent neural networks (RNNs)

    Example: Perform path planning to learn the best possible route to a destination. The Pong example in GitHub uses reinforcement learning.

  • Predict an Output Based on Historical and Current Data

    Applications: Predictive maintenance, financial trading

    Input: Sensor data, timestamped financial data, numeric data

    Common algorithms: Linear regression, decision trees, support vector machines (SVMs), neural networks

    Example: Use real-time sensor data from a motor to predict remaining useful life for rotating machinery. The Similarity-Based Remaining Useful Life Estimation example uses linear regression.

  • Synthesize Images, Signals, and Text

    Applications: Generate failure data, increase size of training set, experiment with audio enhancement

    Input: Images, signal, and text data

    Common algorithms: GANs, autoencoders

    Example: Create images of new types of flowers using existing data. The Train Generative Adversarial Network example uses a GAN with two networks, a Generator to create the images, and a Discriminator that classifies images as “real” or “generated.”

  • Uncover Trends, Sentiments, Fraud, or Threats

    Applications: Natural language processing for safety records, market or medical research, sentiment analysis, cybersecurity, document summarization, recommender systems

    Input: Streaming text data, static text data

    Common algorithms: RNNs, linear regression, SVMs, naïve Bayes, latent Dirichlet allocation (LDA), latent semantic analysis, word2vec

    Example: Determine how many topics are present in text data. The Analyze Text Data Using Topics Models example uses the LDA topic model.

  • Try for Yourself

    Learn how to apply these techniques with MATLAB through free 2-hour tutorials.

    Machine Learning Onramp
    Deep Learning Onramp