CCNA 3 Supplemental Module 1.2

CCNA 3 Supplemental Module 1.2

1. SUPPLEMENTAL MODULE

   • ARTIFICAL INTELIGENCE AND MACHINE LEARNING
     • MACHINE LEARNING
       • CARACTERISTICIS MAIN
         • ABILITY TO LEARN FROM DATA AND IMPROVE PERFORMANCE OVER TIME WITHOUT EXPLIIT PROGRAMMING
       • TYPES OF IAs / ML
         • NARROW IA
           • AS THEY ARE DESIGNED TO PERFORM SPECIFIC TASKS.
           • FOR EXAMPLE, A MACHINE LEARNING MODEL MIGHT BE TRAINED TO RECOGNIZE SPEECH, TRANSLATE LANGUAGES, OR RECOMMEND PRODUCTS, EXCELLING ONLY IN THE SPECIFIC TASK IT WAS DESIGNED FOR.
           • EXAMPLES INCLUDE GOOGLE TRANSLATE AND AMAZON ALEXA.
         • DESCRIPTIVE AI
           • MACHINE LEARNING ENHANCES DESCRIPTIVE AI BY IDENTIFYING PATTERNS, DETECTING ANOMALIES, AND EXTRACTING FEATURES.
           • IT IMPROVES DATA VISUALIZATION, ENABLES AUTOMATED REPORTS, AND PROVIDES MORE ACCURATE AND EFFICIENT DATA ANALYSIS AND SUMMARIZATION.
           • PAST
         • PREDICTIVE AI
           • ANALYZING HISTORICAL DATA TO FORECAST FUTURE EVENTS OR BEHAVIORS,
           • SUCH AS SALES TRENDS IN BUSINESS APPLICATIONS LIKE NETFLIX, SPOTIFY, AND AMAZON.
           • FUTURE
         • GENERATIVE AI
           • GENERATIVE AI USES MACHINE LEARNING, ESPECIALLY MODELS LIKE GANS, TO CREATE NEW DATA SIMILAR TO TRAINING DATA, SUCH AS IMAGES, TEXT, OR MUSIC.
           • GANS USE ONE NETWORK TO GENERATE DATA AND ANOTHER TO EVALUATE IT, IMPROVING REALISM, COMMONLY USED FOR IMAGES IN GAMES AND MOVIES.
         • OTHER EXAMPLES OF ML APPLICATIONS
           • FRAUD DETECTION
             • FINANCIAL INSTITUTIONS USE ML TO DETECT UNUSUAL BEHAVIOR PATTERNS INDICATING FRAUDULENT ACTIVITY IN CREDIT CARD ACCOUNTS OR BANKING OPERATIONS.
           • PERSONALIZED MARKETING
             • COMPANIES USE ML TO ANALYZE CUSTOMER DATA AND CRAFT PERSONALIZED MARKETING STRATEGIES, OFFERING TAILORED PROMOTIONS AND ADVERTISEMENTS.
           • HEALTHCARE PERSONALIZATION
             • ML ALGORITHMS HELP PERSONALIZE PATIENT CARE PLANS BY ANALYZING INDIVIDUAL HEALTH RECORDS AND TAILORING TREATMENTS BASED ON PREDICTIVE OUTCOMES.
           • AUTONOMOUS VEHICLES
             • SELF-DRIVING CARS USE ML TO PROCESS SENSORY DATA, MAKE REAL-TIME DECISIONS, AND LEARN FROM EACH DRIVE TO IMPROVE PERFORMANCE AND SAFETY.
           • SUPPLY CHAIN OPTIMIZATION
             • ML HELPS BUSINESSES PREDICT DEMAND, OPTIMIZE INVENTORY LEVELS, AND ENHANCE SUPPLY CHAIN EFFICIENCY.
           • SPEECH RECOGNITION
             • VIRTUAL ASSISTANTS LIKE GOOGLE ASSISTANT AND APPLE’S SIRI USE ML FOR VOICE RECOGNITION, ALLOWING THEM TO UNDERSTAND AND RESPOND TO SPOKEN COMMANDS.
     • HOW IA LEARNS
       • VARIOUS APPROACHES THROUGH ALGORITHMS CAN LEARN
       • SUPERVISED LEARNING
         • TEACHING AI TO RECOGNIZE A CAT BY IDENTIFYING ATTRIBUTES LIKE POINTY EARS, TAIL, AND WHISKERS IS DIFFICULT DUE TO MANY POSSIBILITIES AND DIFFERENT POSITIONS
         • AI CAN LEARN BY OBSERVING LABELED IMAGES OF CATS IN DIFFERENT SITUATIONS AND CONTEXTS
         • THE MORE IMAGES USED FOR LEARNING, THE MORE ACCURATE THE AI BECOMES
           • THIS IS KNOWN AS SUPERVISED LEARNING
         • SUPERVISED LEARNING TRAINS AI SYSTEMS USING LABELED DATASETS
         • IN NETWORK SECURITY, IT IS USED TO CLASSIFY NETWORK TRAFFIC OR PREDICT THREATS USING DATA LABELED AS "SAFE" OR "MALICIOUS"
           • AI SYSTEMS IDENTIFY PATTERNS AND RELATIONSHIPS BETWEEN DIFFERENT TYPES OF NETWORK ACTIVITIES
         • THE PROCESS INVOLVES PROVIDING INPUT DATA AND EXPECTED OUTPUTS TO TEACH THE ALGORITHM WHAT TO LOOK FOR
       • UNSUPERVISED LEARNING
         • UNSUPERVISED LEARNING DOES NOT RELY ON PRE-LABELED DATA
         • IT IS USED TO FIND USEFUL INSIGHTS IN LARGE AMOUNTS OF DATA
         • VIDEO SERVICES LIKE NETFLIX ANALYZE CUSTOMER VIEWING HABITS TO IDENTIFY SIMILAR USERS AND RECOMMEND MOVIES
         • ALGORITHMS EXPLORE DATA FREELY TO UNCOVER PATTERNS OR GROUPINGS
         • IT CAN IDENTIFY UNUSUAL TRAFFIC PATTERNS THAT DEVIATE FROM NORMAL BEHAVIOR
         • IT CAN FLAG POTENTIAL SECURITY BREACHES WITHOUT PRIOR KNOWLEDGE
         • THE MODEL CAN IDENTIFY ANOMALIES IN THE DATA
           • IT IS USEFUL IN NETWORKING FOR ANOMALY DETECTION
       • REINFORCEMENT LEARNING
         • REINFORCEMENT LEARNING IS A TYPE OF MACHINE LEARNING WHERE AN AGENT LEARNS BY PERFORMING ACTIONS AND RECEIVING REWARDS OR PENALTIES
         • THE GOAL IS TO MAXIMIZE CUMULATIVE REWARDS OVER TIME
         • ALPHAZERO USED REINFORCEMENT LEARNING BY PLAYING MANY GAMES AGAINST ITSELF AND IMPROVING ITS STRATEGIES
         • IT USED A REWARD FUNCTION WITH POSITIVE VALUES FOR WINS AND NEGATIVE VALUES FOR LOSSES
         • REINFORCEMENT LEARNING IS BASED ON TRIAL-AND-ERROR LEARNING
         • THE AI SYSTEM LEARNS FROM THE RESULTS OF ITS ACTIONS
         • THIS HELPS NETWORK SYSTEMS ADAPT AND OPTIMIZE THEIR OPERATIONS
           • IT CAN IMPROVE PERFORMANCE BY ADJUSTING CONFIGURATIONS AND POLICIES BASED ON REAL-TIME FEEDBACK
       • HALLUCINATION AND RAG
         • AI SYSTEMS LEARN FROM LARGE AMOUNTS OF DATA BUT ARE NOT PERFECT AND CAN PRODUCE INCORRECT OR MISLEADING INFORMATION
         • THIS ISSUE IS CALLED "HALLUCINATION" AND OCCURS BECAUSE MODELS RELY ON PATTERNS IN TRAINING DATA RATHER THAN FACTUAL ACCURACY
         • TRAINING DATA MAY BE INCORRECT OR OUTDATED, AND RETRAINING MODELS IS TIME-CONSUMING AND RESOURCE-INTENSIVE
         • RETRIEVAL-AUGMENTED GENERATION (RAG) REDUCES HALLUCINATIONS AND OUTDATED ANSWERS
         • RAG COMBINES A RETRIEVAL SYSTEM WITH A GENERATIVE MODEL
         • THE RETRIEVAL SYSTEM FETCHES RELEVANT INFORMATION, AND THE GENERATIVE MODEL USES IT TO CREATE RESPONSES
         • THIS ENSURES RESPONSES ARE BASED ON REAL-WORLD INFORMATION, IMPROVING ACCURACY AND REDUCING FABRICATION
         • RAG ANCHORS RESPONSES TO ACTUAL CONTENT, MAKING THEM MORE RELIABLE
         • RAG IS USEFUL FOR NAVIGATING COMPLEX TECHNICAL DOCUMENTATION AND EXTRACTING RELEVANT INFORMATION