Page 192 - AI Ver 3.0 class 10_Flipbook
P. 192

Supervised Vs. Unsupervised Learning

              The differences between Supervised and Unsupervised Learning are as follows:
               Aspect                Supervised Learning                         Unsupervised Learning
                                     Uses labelled datasets with input-output    Uses unlabelled datasets without
               Data
                                     pairs (e.g., images labelled as "cat" or "dog"). predefined categories.
                                     To predict the output for new, unseen data   To explore data and find hidden
               Goal
                                     based on the labels provided during training. patterns or groupings.
               Common Techniques     Classification, Regression                  Clustering, Dimensionality Reduction
                                     - Predicting house prices based on features   - Grouping  customers  into  segments
               Examples               like area and location.                     based on their purchase behaviour.

                                     -Classifying emails as "spam" or "not spam".  -Detecting anomalies in network traffic.
                                     Predicts specific labels or values for new   Groups data into clusters or finds
               Output
                                     data.                                       patterns, without specific labels.
              Reinforcement Learning

              Reinforcement Learning is a type of machine learning where a model learns through trial and error to make the
              best decisions in a given environment. It interacts with its surroundings, receiving rewards for correct actions and
              penalties for mistakes, gradually improving over time. The goal is to maximise cumulative rewards by learning from
              experience and adapting to new situations.
              For example, a robot learning to walk starts with random movements. When it takes a correct step, it gets a reward;
              if it falls, it receives a penalty. Over time, the model learns which movements keep it balanced and eventually walks
              efficiently.
              Let us take an example of Reinforcement Learning:

              You show the machine an image of a ball and ask it to predict the object. Initially, the machine predicts it to be a
              globe. Since this is incorrect, it receives negative feedback. The machine then adjusts its understanding and learns
              that the object is not a globe. When the same image is provided again, the machine predicts it to be an orange.
              This is also incorrect, so it receives negative feedback again.







                                  Reinforcement Model     Globe                   Reinforcement Model    Orange



              Finally, after  multiple attempts and  adjustments, the machine predicts the object correctly  as a ball. Positive
              feedback is given, and the machine has successfully learned to identify the ball.




                                                          Reinforcement Model      Ball











                    190     Touchpad Artificial Intelligence (Ver. 3.0)-X
   187   188   189   190   191   192   193   194   195   196   197