Understanding Machine Learning --- Chapter 2.pptx

Understanding Machine Learning --- Chapter 2.pptx

• 1.
  SOFTWARE CONSTRUCTION AND DEVELOPMENT BY SANARIZWAN CHAPTER 2 UNDERSTANDING MACHINE LEARNING
• 2.
  CONTENTS • Construct ofAI and ML • Types of Machine Learning • Applications of Machine Learning • Challenges and Future Directions • AI techniques and Algorithms • Supervised Learning Algorithms • Unsupervised Learning Algorithms • Reinforcement Learning Algorithms • Deep Learning Algorithms • LLM, BERT, XLNet, T5 2
• 3.
  CONSTRUCT OF AIAND ML 3 Artificial intelligence development involves teaching machines to learn and to use those lessons to generate certain outcomes. While artificial intelligence involves learning and delivering expected outcomes, the learning phase is referred to as machine learning (ML). Learning is accelerated through spider crawlers (like Google crawlers), which enable machines to learn from the information that is available on public forums, apart from what is manually fed to them. ML is the driver that makes AI relevant because the knowledge the machines receive is enriched by the day, and the outcomes are dynamic based on the current situation.
• 4.
  4 THE STRUCTURE • Machinelearning is a subset of the artificial intelligence. • Machine learning is the process or activity where machines learn from the data provided to them. This learning improves over time, with new datasets replacing existing ones, so machine learning is a continuous process. • There are several learning models, including deep learning, computer vision, and reinforcement learning, among others. • Generative AI (Gen AI) is a technology that can produce content, including videos, images, text, and audio. Big Data plays a pivotal role in shaping AI and ML. The amount of clean data that is fed into the machines makes the AI smarter. Machines can identify patterns and trends and predict the future course of action.
• 5.
  5 TYPES OF MACHINELEARNING
• 6.
  6
• 7.
  7
• 8.
  8
• 9.
  9 APPLICATIONS OF ML Machinelearning finds focus in areas where decision making is done on the back of data analysis. All sectors and areas of technology use a wide variety of data for decision making. A good leader relies on the data, which impacts decision making. Machine learning does the same. It reads and analyzes data that will help it determine the best course of action, and depending on the processing engines used, decisions are made accordingly. • Shopping and content recommendations • Reading X-rays, ECGs, and MRIs • Helping the finance and investment sectors • Translating audio in real-time for seamless conversations • Assessing people’s skills and developing dynamic training content
• 10.
  10 CHALLENGES AND FUTURE DIRECTIONS
• 11.
  11 KEY AI ALGORITHMS •Artificial intelligence algorithms are several hundred/thousand times more complex. They are not merely trying to find possible outcomes but are building a system to think like humans. • To provide machines with human-like intelligence and perform analysis and processing, several artificial intelligence techniques and algorithms are employed, each suited to different types of problems and data. • Every algorithm has a specific purpose and leads to a range of outcomes. • Integrating AI techniques and algorithms into DevOps and MLOps workflows enables automation, optimization, and decision making in various aspects of software development, deployment, and operations. • These techniques help streamline processes, improve efficiency, and enhance the reliability of software systems in production environments.
• 12.
  12 SUPERVISED LEARNING
• 13.
  13 SUPERVISED LEARNING
• 14.
  14
• 15.
  15
• 16.
  16
• 17.
  17
• 18.
  18
• 19.
  19 LARGE LANGUAGE MODELS(LLM) • A Large Language Model (LLM) refers to a machine that understands our natural language, in the manner that we speak, including the words, grammar, and style. • AI systems understanding our language is one part of the equation, LLMs are trained and designed to produce output of human-like text at scale. • AI models have undergone massive lengths of training with vast amounts of text data. • They use sophisticated algorithms to learn the shades of the language, including grammar, syntax, semantics, and context. • The primary objective of an LLM is to perform numerous natural language processing (NLP) tasks, such as text generation, translation, summarization, sentiment analysis, and question answering. • OpenAI’s GPT (Generative Pre-trained Transformer) is a popular LLM, and there are multiple models, such as GPT-2 and GPT-3. • They have been trained for several days, with massive datasets containing diverse text, generally gathered through web spiders from the Internet. • The model is quite mature and is capable of generating contextually relevant text depending on the given prompt or input. • LLMs are used in various sectors and have a wide range of applications across industries, including content generation, customer service automation, language translation, and virtual assistants, among others. • They have taken AI capabilities to greater heights in understanding and generating natural language and they are paving the way to innovative solutions in various fields.
• 20.
  20 • By leveragingthe capabilities of different types of LLMs in DevOps contexts, organizations can automate routine tasks, improve communication and collaboration, and enhance overall productivity and efficiency in software development and operations. • LLMs come in various types, each with its own architecture, training methods, and applications. The following sections describe some common types of LLMs. • Transformer-Based Models • BERT (Bidirectional Encoder Representations from Transformers) • XLNet • T5 (Text-To-Text Transfer Transformer) • BERT-Based Models for Domain-Specific Tasks • Multilingual LLMs LARGE LANGUAGE MODELS (LLM)
• 21.
  21
• 22.
  22
• 23.
  23
• 24.
  24
• 25.
  25
• 26.
  26
• 27.
  THANK YOU Sana Rizwan AssistantProfessor Department of Computer Science, COMSATS University Islamabad, Lahore Campus Email: sana@cuilahore.edu.pk 27