Perception: Natural Language Processing and Computer Vision

Content

• Stochastic fundamentals: Markov networks, Markov random fields, dynamic Bayesian networks, query and query
  response algorithms, sampling methods, learning methods with incomplete data (expectation-maximization: EM, Baum-Welch method), PAC learning principle
• Probabilistic language models, topic models, latent Dirichlet allocation (LDA), thematic developments over time represented with dynamic Bayesian networks
• Transformation networks as probabilistic models, training methods for convolutional and transformation networks, application-specific training through fine-tuning (deep and shallow), integration of special (symbolic) problem solvers
  in GPTs, differential programming
• Analysis of videos: Object detection with transformation network architectures (YOLO)
• Probabilistic computational networks (e.g. with applications in image processing), query answering and scalability, transformation of probabilistic models to probabilistic computational networks, control of large language models with probabilistic models
• Generation of relevant new objects to simplify the process of finding solutions to problems (e.g. AlphaGeometry,
  FunSearch)
• Generative modeling of data (e.g. images):
• Generation of images and videos: Variational autoencoder with vector quantization (DALL-E), denoising diffusion,
  outpainting and inpainting
• Construction of complex probability distributions through a series of invertible transformations: Normalizing Flows, Combination with Probabilistic Computational Networks, Generating Adversarial Networks (GANs)