Why this technology is relevant and what the scope.
Computers transformed the way we interact with things in our daily life. We have a software in almost everything we use, from toaster to our cars. We are witnessing an age where it is being taken to the next level by integrating AI to our daily life things and thereby solving a vast majority of problems we face in medical science to space exploration. AI is the next big thing.
One use machine learning or AI to build softwares that do human level tasks that's hard to write as code
AI will create 133 million jobs by 2025.
Machine Learning is creating or learning models from input data and produce predictions on never before seen data.
https://encrypted-tbn0.gstatic.com/images?q=tbn%3AANd9GcRhnT-A-5jO2W-WGCvty8EvVS7LlZ0ezprPlshhBLn7T3XYduF8
Creating/learning of ML models is called training. We feed the input data and target output and run the algorithm in a loop till the algorithm learn the underlying patterns in the data. This takes comparatively more time and computation resources than Inference.
Inference means running trained model with never before seen data and making predictions. After training the model we give an input to a trained model and we get the desired output.
Supervised Learning is Machine Learning with Labelled Dataset. Labelled dataset will have features and labels in the dataset.
In the below image, images of ducks, rabbit and human are features.
In the image below, Duck
and Not duck
are labels.
In the training phase we feed the features and labels to the supervised learning algorithm and the algorithm will learn the patterns associated with the features and corresponding labels.
In inference phase we pass an input image to the trained predictive model and the model will output the corresponding label based on the patterns learnt.
In unsupervised learning, model will learn the underlying pattern from the dataset without manual supervision.
The dataset is fed to the algorithm without any labels. Algorithm will automatically find patterns and divide dataset into different different categories.
We input a data and the model will add the data to corresponding group/cluster based on the pattern learnt
Agent learn the best action patterns for a purpose by doing several iterations of different actions in the environment and gaining rewards.
For example, RL can be used to train a model for playing flappy birds game.
https://encrypted-tbn0.gstatic.com/images?q=tbn%3AANd9GcQFR5ikyS0lbsX7V7dgai6LpeNlDpxt-XR82Vb4yurR5SQAMCb_
The environment is the world that the agent lives in and interacts with. For example the environment of flappy bird.
Agent is the actor in the environment. For example, the bird in flappy bird.
Agent apply an action to the environment (for example up arrow at a particular frame in flappy bird) and move to a new state from the current state.
A reward is given to the agent, which indicate how good the new state is. The goal of the agent is to maximise its cumulative reward.
Reinforcement learning methods are ways that the agent can learn behaviors to achieve its goal.
https://www.youtube.com/watch?v=kopoLzvh5jY
<aside> 👶 Level: Beginner
</aside>
<aside> 💼 Career Path: ‣
</aside>
A Neural Network in 11 lines of Python (Part 1)
Gopikrishnan Sasikumar
Mohammed Faiz K
Muhammad Hafil
MUHAMMAD ASLAM PT