Page 178 - Robotics and AI class 10
P. 178

Whether the ball was of the same  colour at the back or not? Or was it just  a hemisphere? If we reduce the
        dimensions further, more and more information will get lost. So, we use Dimensionality Reduction here which
        reduces the dimensions and makes it sensible data.

               Z                                    Y                            Y




                           Y




                                      X                                     X                            Z

        Reinforcement Learning
        Reinforcement learning is a type of learning based approach where a machine learning algorithm enables an agent
        (machine with an intelligent code) to learn in an environment to find the best possible behaviour or path it should
        take by performing certain actions that maximize the total cumulative reward of the agent.
        In this learning approach the agent  learns automatically by using hit and trial methods or through its own
        experience using rewards and penalties. Each action performed by an agent gives reward for correct move and
        it signals positive feedback. For wrong move it generates negative feedback and gets punishment and a penalty.

        The agent explores the environment by interacting with it freely so that it is able to improve the performance by
        getting the maximum positive rewards.
        The best applications of reinforcement learning are self driving cars, robotics and a variety of video games available
        these days.


                 Evaluation


        Evaluation is seen as the end of the Project cycle. It is an important step where the AI model is evaluated for its
        efficiency and accuracy. It enables continuous improvement of the model to achieve the project goals. The model
        must be tested with varied data to ensure that the results are satisfactory. Model is tested with the testing data
        after each stage of the AI project cycle. Final evaluation must be done to check the overall functioning of the
        model. Once the model is evaluated it must be deployed.
        The general steps for evaluation and testing of the AI model would be as follows:

        Step 1:  The testing data is fed to the algorithm.
        Step 2:  The model processes the data as per the data it was trained on.

        Step 3:  Model does the prediction.
        Step 4:  The prediction is compared with the test data set value.

        Step 5:  Model testing continues with more test data sets.
        Step 6:  The prediction value of the testing data is compared with the Actual values.

        Step 7:  If the predicted value matches or is similar or close to the actual value, the model is considered working
                accurately, and no changes are required and is ready for deployment. If the predicted value is off and not
                close to the actual data, then the model is reviewed and changed, or trained further with more data, to
                get the prediction more accurate.


              176     Touchpad Robotics & Artificial Intelligence-X
   173   174   175   176   177   178   179   180   181   182   183