With the development of Artificial Intelligence and Machine Learning, the field of Reinforcement learning is in great demand, due to its self-learning abilities, here we are proposing some of the examples of Reinforcement learning or Reinforcement learning examples in a little detailed format.
- What is Reinforcement learning?
- How does Reinforcement learning work?
- Why we learn Reinforcement learning
- Practical Reinforcement learning examples:
- 1) Reinforcement learning in Training Neural Networks for classification:
- 2) Reinforcement learning in Making autoplay game of pong:
- 3) Reinforcement learning in E-commerce (Online Recommendation):
- 4) Reinforcement learning in Trading:
- 5) Reinforcement learning in Managing Marketing Strategies:
- 6) Reinforcement learning in Reduction in Energy Consumption:
- 7) Reinforcement learning in Optimisation of Bidding system & Gambling:
- 8) Reinforcement learning in Industrial Machine Training:
- 9) Reinforcement learning in Text Mining:
- 10) Reinforcement learning in Manufacturing Quality products:
- 11) Reinforcement learning in Self Driving Cars:
- 12) Reinforcement learning in NLP (Natural Language Processing):
- 13) Reinforcement learning in Healthcare advancements:
- 14) Reinforcement learning in News, Feeds & Article Recommendation:
- 15) Reinforcement learning in Gaming Optimisation:
- 16) Reinforcement learning in Optimised & Targeted Advertising:
- 17) Reinforcement learning in Robotic Manipulation:
- 18) Reinforcement Learning optimized applications in the engineering field:
- 19) Reinforcement learning in Image Processing:
What is Reinforcement learning?
Right from Self-playing chess to Tetris, self-moving robotics, and many other fields that require constant improvement Reinforcement learning is always in limelight. The field of Reinforcement learning has exploded due to the development of Deep learning, a better understanding of neural networks, a breakthrough in machine learning and AI.
Reinforcement learning is a part of Machine Learning, it’s an area of Machine Learning that works on constant self-development algorithms in simple terms we can say due to Reinforcement learning we can develop a software, program, algorithm, system, etc that has the ability to develop itself in terms of speed, accuracy, precision.
In technical terms, we can say Reinforcement learning is a training part of Machine Learning that deals with the decision in a sequential manner with the help of a trial and error method that works on a reward system.
Basically in Reinforcement learning works on trial and error methods to come up with the most accurate answer that requires lots of computing power and advance programming skills. In other words, we can say it is one of the backbones of Artificial Intelligence, which helps AI to develop on its own.
With the help of reinforcement learning, we can combine artificial Neural Networks to create Deep Reinforcement learning, which can basically work or act as Artificial Intelligence that can work on its own, or can think on its own.
How does Reinforcement learning work?
Since to know how Reinforcement learning works we have to know about how Supervised Learning works, and supervised learning is a part of deep neural nets what we call Deep learning.
Let’s take a game of chess in this if we apply Supervised Learning we have to use an actual human player and note down his moments and statistics and convert that raw data into distilled datasets with the help of data mining, and then we can configure the self-playing game of chess, but here’s a drawback of that it cannot learn by itself an actual person has to train in Supervised Learning.
But this is not the case with Reinforcement learning, it works on a reward mechanism, it tries multiple ways to play in chess, and when it wins it considers as a reward, and when it loses it considers nothing, by using this technique it constantly improves its own algorithm.
How Reinforcement learning works Example
Let’s take an example of a cat as an example of Reinforcement learning working, if we want to train a cat to do some specific task we have to give some kind of reward, that will encourage the cat to do more things as we say.
Here the fish act’s as a reward point that indirectly sends signals to the cat that it’s a good thing to do, which is to obey our commands in our cat scenario. In this when the cat completes the task assigned by us we give her a fish (reward), which releases a little bit of dopamine in the cat and links happiness and reward for obeying the command.
Like-wise in our case Reinforcement learning works the same, unlike supervised Learning in which if there is no reward related to the output, but in the case of Reinforcement learning, we are adding the reward point as the indicator to continue the task, we are not providing any kind of fish to the Reinforcement learning model, instead of in case of Reinforcement learning the fish acts as an indicator to continue the process otherwise abort the process.
If you want to know about ARTIFICIAL SUPER INTELLIGENCE check out our latest blog.
Why we learn Reinforcement learning
From the amazing results and vintage Atari games to stunning breakthroughs in robotic arm manipulation and even beating professional players at 1v1 dota the field of reinforcement learning has literally exploded in recent years.
Ever since the impressive breakthrough on the image classification challenge in 2012 the successes of supervised deep learning have continued to pile up and people from many different backgrounds have started using deep neural nets to solve a wide range of new tasks including how to learn intelligent behavior in complex dynamic environments.
We all know Artificial Intelligence is the future, and many companies are putting millions of dollars into it, and Machine Learning, Deep Learning, Neural Networks, and Reinforcement learning are the stepping stones towards AI.
Due to this, there is a big surge of competition on making better Machine Learning and Artificial Intelligence models, which is in high demand, due to which many companies and programmers are urging to develop better Reinforcement learning models that can be used in our day-to-day purpose.
So if you want to make a carrier in Artificial Intelligence you have to master the application of Reinforcement learning precisely.
One of the reasons to learn AI is in Business, checkout “18 Future Scope of Artificial Intelligence in Business“
Practical Reinforcement learning examples:
As Reinforcement learning is a developing field there are many places where RL is applicable, many companies are secretly applying Reinforcement learning models on their business models for better performance and greater profitability.
So these are the Examples where Reinforcement learning has been used / Reinforcement learning uses / Reinforcement learning examples.
1) Reinforcement learning in Training Neural Networks for classification:
Neural Classifications are one of the major aspects of Reinforcement learning development, in this session we are comparing Reinforcement learning with supervised Learning.
We are taking an example of image classification, in this, we are taking the example of cat identification, that whether the image of the animal is a cat or a dog or something else.
I want to introduce you guys to the whole subfield in machine learning that’s called reinforcement learning which i think is one of the most promising directions to actually get to very intelligent robotic behavior so in the most common machine learning applications.
people use what we call supervised learning and this means that you give an input to your neural network model but you know the output that your model should produce and therefore you can compute gradients using something like the back propagation algorithm to train that network to produce your outputs
But in the case of Reinforcement learning, we treat the correct answer as a reward point so that it can continue its actions or else it can stop.
2) Reinforcement learning in Making autoplay game of pong:
Let’s take an example of pong, in which we are taking actual working of RL model.
so imagine you want to train a neural network to play the game of pong what you would do in a supervised setting is you would have a good human gamer play the game of pong for a couple of hours and you would create a data set where you log all of the frames that that human is seeing on the screen as well as the actions that he takes in response to those frames.
so whatever is pushing the up arrow or the down arrow and we can then feed those input frames through a very simple neural network that at the output can produce two simple actions it’s either going to select the up action or the down action
and by simply training on the data set of the human gameplay using something like backpropagation we can actually train that neural network to replicate the actions of the human gamer.
but there are two significant downsides to this approach so on the one hand if you want to do supervised learning you have to create a data set to train on which is not always a very easy thing to do and on the other hand if you train your neural network model to simply imitate the actions of the human player well then by definition your agent can never be better at playing the game of pong than that human gamer.
For example, if you want to train a neural net to be better at playing the game of pong and the best human then by definition we can’t use supervised learning, so is there a way to have an agent learn to play a game entirely by itself well, fortunately, there is and this is called reinforcement learning.
So the framework and reinforcement learning are actually surprisingly similar to the normal framework in supervised learning.
so we still have an input frame we run it through some neural network model and the network produces an output action we either up or down but the only difference here is that now we don’t actually know the target label.
so we don’t know in any situation whether we should have gone up or down because we don’t have a data set to train on and in reinforcement learning the network that transforms input frames to output actions is called the policy Network.
now one of the simplest ways to train a policy network is a method called policy gradients.
So the approach in policy gradients is that you start out with a completely random network you feed that network a frame from the game engine it produces a random up with action you know either up or down you send that action back to the game engine and the game engine produces the next frame and this is how the loop continues and the network.
in this case it could be a fully connected network but you can obviously apply convolutions there as well and now in reality the output of your network is going to consist of two numbers the probability of going up and the probability of going down and what you will do while training is actually sample from the distribution
o that you’re not always going to repeat the same exact actions and this will allow your agent to sort of explore the environment a bit randomly and hopefully discover better rewards and better behavior
now importantly because we want to enable our agent to learn entirely by itself the only feedback that we’re gonna give it is the scoreboard in the game so whenever our agent manages to score a goal it will receive a reward of +1 and if the opponent scored a goal then our agent will receive a penalty of -1.
the entire goal of the agent is to optimize its policy to receive as much reward as possible so in order to train our policy network the first thing we’re gonna do is collect a bunch of experience so you’re just gonna run a whole bunch of those game frames through your network select random actions feed them back into the engine and just create a whole bunch of random pong games.
and now obviously since our agent hasn’t learned anything useful yet it’s gonna lose most of those games but the thing is that sometimes our agent might get lucky sometimes it’s going to randomly select a whole sequence of actions that actually lead to scoring a goal and in this case our agent is going to receive a reward and a key thing to understand is that for every episode regardless of whether we want a positive or a negative reward
we can already compute the gradients that would make the actions that our agents has chosen more likely in the future and this is very crucial and so what policy gradients are going to do is that for every episode where we’ve got a positive reward we’re going to use the normal gradients to increase the probability of those actions in the future
but whenever we got a negative we’re gonna apply the same gradient but we’re gonna multiply it with minus one and this minus sign will make sure that in the future all the actions that we took in a very bad episode are going to be less likely in the future and so the result is that while training our policy network the actions that lead to negative rewards are slowly going to be filtered out and the actions that leads to positive rewards are going to become more and more likely
so in a sense, our agent is learning how to play the game of pong.
3) Reinforcement learning in E-commerce (Online Recommendation):
Customer satisfaction is the most important thing in online business, with the help of Reinforcement learning we can provide customization in choices according to the taste of each individual.
A customized opinion system that can be powered with the help of Machine Learning, there are many factors that depend on which product a customer should buy like his budget range, favorite brand, color, etc.
Reinforcement learning is being extensively used in online recommendation systems, right from buyer to retailer, types of products to a range of availabilities.
Recently Chinese Nanjing University came together with Alibaba Group to build a reinforcement learning algorithm that uses an online recommendations system for better customer service and satisfaction.
4) Reinforcement learning in Trading:
Trading is a risky field and requires lots of experience, with the help of Reinforcement learning we can train bots to work as an online trader, the reason why we are using Reinforcement learning in this sensitive field
In current times the stock market is one of the hottest and booming fields where Artificial Intelligence & Reinforcement learning can be applied. Synthetic trading of Algorithmic trading is an old field of trading cryptocurrencies and stocks, the main reason behind these applications is because of high accuracy and greater chances of profitability.
Many companies like Pit.ai have been implementing Reinforcement learning in stock & crypto trading.
5) Reinforcement learning in Managing Marketing Strategies:
For many companies Marketing is the key to success and for many years many corporate giants are investing millions of dollars in marketing campaigns and their strategies.
Creating and managing dynamic marketing strategies is one of the examples of Reinforcement learning, RL helps to track down customer satisfaction points that create huge data sets that can be beneficial for profitable marketing strategies.
Recently Researchers from the University of New York formed a unique type of algorithm known as “Inverse Reinforcement Learning”, which is capable of predicting customer choices, decisions, reactions, and behavior by simulation of their change.
6) Reinforcement learning in Reduction in Energy Consumption:
Constant consumption of energy is one of the major future issues, and it is considered to be one of the major issues because the demand is rapidly increasing and the supply is not enough.
Due to constant consumption, there are many ways to save energy, and with the help of Reinforcement learning we can make the process much accurate and better as compared to humans.
There are many ways to save or reduce energy consumption with the help of Machine Learning and Reinforcement learning, like choosing the optimal methods to energy transfer to production.
Recently tech giant Google is working on such a project, that uses DeepMind AI to find out the most optimal method for cooling their data-centers and other services.
7) Reinforcement learning in Optimisation of Bidding system & Gambling:
Bidding and Gambling is all game of Dopamine, we feel good when we win the lottery or grab a deal in bidding wars, here comes one of the most cunning examples of Reinforcement learning, and i.e bidding & gambling.
In this, we are taking an example of online ads bidding system, as shown in the figure Reinforcement learning works on customer reviews and feedbacks and analyze them as +ve points and stores them in its dataset rather than using previous old reviews, it works on time and makes on-time profitable results for advertisers.
Likewise, it also works in other systems like gambling, it recognises the best gambling pool and other criteria and based on that it customizes its algorithm for maximum profitability.
8) Reinforcement learning in Industrial Machine Training:
Industrial Training is a very tedious and stressful task, with the help of Reinforcement learning we can make industrial teaching much more accurate and fast as well.
9) Reinforcement learning in Text Mining:
Text mining is one of the most important applications of NLP (natural language processing), one of the most famous cloud computing company salesforce uses text mining.
The system of salesforce produces well-structured & readable text and can convert them into long summaries as well.
The application of text mining is on generating well-optmised articles, blogs, news, etc.
10) Reinforcement learning in Manufacturing Quality products:
Manufacturing of quality products depends on customer reviews and feedbacks, with the help of Reinforcement learning we can create more similar types of products that performed better in sales and have less customer negative feedback.
Positive feedback and reviews indicate +ve point to the Reinforcement learning algorithm, which we can use to make better and better products.
11) Reinforcement learning in Self Driving Cars:
According to many researchers, Self-driving car is the future, companies like Tesla are investing heavily in developing fully automatic self driving cars.
Artificial Intelligence, Image Recognition, Machine Learning, Reinforcement learning, Deep learning, Neural Nets, etc are some of the fields that are applied in the field of Self-driving cars.
With the help of AI & Reinforcement learning algorithm, we can make a car drive itself without any human help.
Since the topic is huge, I suggest you visit our latest blog on how tesla’s Autopilot car works with the help of AI, ML, & deep learnings.
If you want to learn more about Tesla AI & their Autopilot AI have a look at “WHAT IS TESLA AI & HOW TESLA’S AUTOPILOT AI WORKS?”
12) Reinforcement learning in NLP (Natural Language Processing):
Reinforcement learning can be used in NLP for many different purposes such as text summarization, answering questions, machine language translation, etc.
In a research paper published by Salesforce, in this paper the authors Romain Paulus, Caiming Xiong & Richard Socher they proposed that how RL is able to convert long text answers or answering the questions.
The method works by selecting the specific sentences that are related to the questions and answers from the document.
In this paper “abstractive text summarization in this paper” they proposed a combination of supervised and reinforcement learning that can be used in text summarization. The main goal of this research paper is that they used an RNN-based encoder and decoder in longer documents,
Machine Translation is also practiced by many researchers, recently authors from the University of Colorado and the University of Maryland, who propose a reinforcement learning-based approach to simultaneous machine translation.
The amazing thing about this research is that with the help of the Reinforcement learning algorithm has the ability to trust its prediction by waiting for the response and then simulates its reply accordingly.
Many researchers from Stanford University, Ohio State University, and Microsoft Research have fronted Deep RL for use in dialogue generation.
This research has the ability to form an advanced Chatbot assistant with the help of Advance Deep Reinforcement learning.
13) Reinforcement learning in Healthcare advancements:
Healthcare is a huge department by itself, so it’s not currently possible to cover all the topics of examples of Reinforcement learning in healthcare, but still, we are trying to perform as much we can.
With the help of reinforcement learning, we are able to predict the related health issues of a specific patient, by scanning and analyzing his past medical treatments and diseases, we can also optimally recommend his health policies related to his previous treatments.
The following image will be helpful for better understanding.
With the help of RL, we can decrease the error causing and can save many many lives as possible.
14) Reinforcement learning in News, Feeds & Article Recommendation:
Writing newsletters, blogs, articles, etc are one of the specialties of reinforcement learning, with the development in RL algorithms many media companies are using text automation that works on NLP and reinforcement learning.
We all saw that google and many applications show news according to our taste, this is because they are applying Machine Learning algorithms for news recommendation in our feeds, Auto news recommendation is the best reinforcement learning example that is available in the market.
In a paper published by Pennsylvania State University, in this research, the authors Guanjie Zheng, Fuzheng Zhang, Zihan Zheng, Yang Xiang demonstrated that how reinforcement learning can be applied in news recommendation, have a look at “Deep Reinforcement learning for news recommendation“.
15) Reinforcement learning in Gaming Optimisation:
Optimizing Games with the help of Reinforcement learning is the best way to learn and understand RL & game mechanics, there’s a game named as AlphaGo Zero in which the developers applied Reinforcement learning algorithms.
They trained their Reinforcement learning model to play the game by itself, and surprisingly after few weeks the RL bot was able to beat the world champion Ke Jie, here are some of the pics.
As the match ended Ke Jei stated, “This Summit is one of the greatest matches that I’ve had. I believe it’s actually one of the greatest matches in history. – KE JIE”
After that, the Reinforcement learning bot also played the whole group of Chinese professionals at once and won against all the odds.
To learn more visit “AlphaGo China“
16) Reinforcement learning in Optimised & Targeted Advertising:
Many tech giants like Facebook, Google, Twitter, Instagram uses Reinforcement learning in their Advertising systems that target specific customers that are related to a specific niche.
These companies use Real-time Bidding for their ads, we all know in today’s competitive marketing field targeting the right customer is a very crucial task to gain maximum profit from little investment, hence RL plays an important role in marketing.
Recently a Research paper published by Alibaba Group, University College London, Shanghai JiaoTong University on “Real-Time Bidding with Multi-Agent Reinforcement Learning” in which they demonstrated that with the help of Reinforcement learning how we can target specific customer and also demonstrated how the algorithm works.
Here are some of the images from that research.
17) Reinforcement learning in Robotic Manipulation:
The reason why we use Robotics is because of its accuracy, precision, speed, working 24×7, and much more, but have you ever thought that we program this robot with our experience and skills, what if these robots learn by themselves?
Here comes the actual application of Reinforcement learning in robotics, where we can train robotic arms, legs to do specific tasks by themselves, with the help of trial and error methods we can train them until they acquire perfection in themselves.
Here are some of the demo images of Example of Reinforcement Learning in Robotics
18) Reinforcement Learning optimized applications in the engineering field:
Engineering is one of the widest fields where Reinforcement Learning can be applied, recently Facebook has developed “Horizon” (The first open-source reinforcement learning platform for large-scale products and services) from which the platform can use RL on large scale in the application.
Facebook’s Horizon can be used for,
- deliver more meaningful notifications to users
- to personalize suggestions
- optimize video streaming quality.
Horizon is capable of handling production-like concerns such as:
- distributed learning
- feature normalization
- deploying at scale
- serving and handling datasets with high-dimensional data and thousands of feature types.
19) Reinforcement learning in Image Processing:
Image Processing is a constantly involving field, with the new advancements in Image recognition systems, AI & ML libraries like OpenCV is in great advancement and in greater demand.
OpenCV is a library of programming functions mainly aimed at real-time computer vision, that relies on the Reinforcement Learning algorithm for constant improvement and development with the help of the trial-n-error method.
If you want to learn more about Reinforcement learning in details have a look at “23 BEST BOOKS TO LEARN REINFORCEMENT LEARNING FOR BEGINNERS TO PROFESSIONAL”