Artificial Intelligence
Artificial Intelligence
AI uses predictive analytics, NLP and Machine Learning to recommend relevant searches to you.
These recommendations are based on data that Google collects about you, such as your search
history, location, age, etc. Thus, Google makes use of AI, to predict what you might be looking for.
• Reactive Machines AI: Based on present actions, it cannot use previous experiences to form current
decisions and simultaneously update their memory.
Example: Deep Blue
• Limited Memory AI: Used in self-driving cars. They detect the movement of vehicles around them
constantly and add it to their memory.
• Theory of Mind AI: Advanced AI that has the ability to understand emotions, people and other things
in the real world.
• Self Aware AI: AIs that posses human-like consciousness and reactions. Such machines have the
ability to form self-driven actions.
• Artificial Narrow Intelligence (ANI): General purpose AI, used in building virtual assistants like Siri.
• Artificial General Intelligence (AGI): Also known as strong AI. An example is the Pillo robot that
answers questions related to health.
• Artificial Superhuman Intelligence (ASI): AI that possesses the ability to do everything that a
human can do and more. An example is the Alpha 2 which is the first humanoid ASI robot.
• Machine Learning: It’s the science of getting computers to act by feeding them data so that they can
learn a few tricks on their own, without being explicitly programmed to do so.
• Neural Networks: They are a set of algorithms and techniques, modeled in accordance with the
human brain. Neural Networks are designed to solve complex and advanced machine learning
problems.
• Robotics: Robotics is a subset of AI, which includes different branches and application of robots.
These Robots are artificial agents acting in a real-world environment. An AI Robot works by
manipulating the objects in it’s surrounding, by perceiving, moving and taking relevant actions.
• Expert Systems: An expert system is a computer system that mimics the decision-making ability of
a human. It is a computer program that uses artificial intelligence (AI) technologies to simulate the
judgment and behavior of a human or an organization that has expert knowledge and experience in
a particular field.
• Fuzzy Logic Systems: Fuzzy logic is an approach to computing based on “degrees of truth” rather
than the usual “true or false” (1 or 0) boolean logic on which the modern computer is based. Fuzzy
logic Systems can take imprecise, distorted, noisy input information.
• Natural Language Processing: Natural Language Processing (NLP) refers to the Artificial
Intelligence method that analyses natural human language to derive useful insights in order to solve
problems.
The Q-learning is a Reinforcement Learning algorithm in which an agent tries to learn the optimal
policy from its past experiences with the environment. The past experiences of an agent are a
sequence of state-action-rewards:
In the above state diagram, the Agent(a0) was in State (s0) and on performing an Action (a0), which
resulted in receiving a Reward (r1) and thus being updated to State (s1).
Deep learning imitates the way our brain works i.e. it learns from experiences. It uses the concepts
of neural networks to solve complex problems.
• Input Layer: This layer receives all the inputs and forwards them to the hidden layer for analysis
• Hidden Layer: In this layer, various computations are carried out and the result is transferred to the
output layer. There can be n number of hidden layers, depending on the problem you’re trying to
solve.
• Output Layer: This layer is responsible for transferring information from the neural network to the
outside world.
• Deep Learning is based on the basic unit of a brain called a brain cell or a neuron. Inspired from a
neuron, an artificial neuron or a perceptron was developed.
• A biological neuron has dendrites which are used to receive inputs.
• Similarly, a perceptron receives multiple inputs, applies various transformations and functions and
provides an output.
• Just like how our brain contains multiple connected neurons called neural network, we can also have
a network of artificial neurons called perceptron’s to form a Deep neural network.
• An Artificial Neuron or a Perceptron models a neuron which has a set of inputs, each of which is
assigned some specific weight. The neuron then computes some function on these weighted inputs
and gives the output.
• The simplest form of ANN, where the data or the input travels in one direction.
• The data passes through the input nodes and exit on the output nodes. This neural network may or
may not have the hidden layers.
• Here, input features are taken in batch wise like a filter. This will help the network to remember the
images in parts and can compute the operations.
• Mainly used for signal and image processing
• Works on the principle of saving the output of a layer and feeding this back to the input to help in
predicting the outcome of the layer.
• Here, you let the neural network to work on the front propagation and remember what information it
needs for later use
• This way each neuron will remember some information it had in the previous time-step.
Autoencoders
• These are unsupervised learning models with an input layer, an output layer and one or more hidden
layers connecting them.
• The output layer has the same number of units as the input layer. Its purpose is to reconstruct its own
inputs.
• Typically for the purpose of dimensionality reduction and for learning generative models of data.
Q11. What are Bayesian Networks?
A Bayesian network is a statistical model that represents a set of variables and their
conditional dependencies in the form of a directed acyclic graph.
On the occurrence of an event, Bayesian Networks can be used to predict the likelihood that any one of several
possible known causes was the contributing factor.
For example, a Bayesian network could be used to study the relationship between diseases and symptoms.
Given various symptoms, the Bayesian network is ideal for computing the probabilities of the presence of
various diseases.
Q12. Explain the assessment that is used to test the intelligence of a machine.
In artificial intelligence (AI), a Turing Test is a method of inquiry for determining whether or not a computer
is capable of thinking like a human being.
Artificial Intelligence Intermediate Level Questions
Q1. How does Reinforcement Learning work? Explain with an example.
1. An agent
2. An environment
• The environment is the setting that the agent is acting on and the agent represents the RL algorithm.
• The RL process starts when the environment sends a state to the agent, which then based on its
observations, takes an action in response to that state.
• In turn, the environment sends the next state and the respective reward back to the agent. The agent
will update its knowledge with the reward returned by the environment to evaluate its last action.
• The loop continues until the environment sends a terminal state, which means the agent has
accomplished all his tasks.
To understand this better, let’s suppose that our agent is learning to play counterstrike. The RL process can be
broken down into the below steps:
1. The RL Agent (Player1) collects state S⁰ from the environment (Counterstrike game)
2. Based on the state S⁰, the RL agent takes an action A⁰, (Action can be anything that causes a result i.e.
if the agent moves left or right in the game). Initially, the action is random
3. The environment is now in a new state S¹ (new stage in the game)
4. The RL agent now gets a reward R¹ from the environment. This reward can be additional points or
coins
5. This RL loop goes on until the RL agent is dead or reaches the destination, and it continuously outputs
a sequence of state, action, and reward.
Q2. Explain Markov’s decision process with an example.
• Set of actions, A
• Set of states, S
• Reward, R
• Policy, π
• Value, V
To briefly sum it up, the agent must take an action (A) to transition from the start state to the end
state (S). While doing so, the agent receives rewards (R) for each action he takes. The series of
actions taken by the agent, define the policy (π) and the rewards collected define the value (V). The
main goal here is to maximize rewards by choosing the optimum policy.
To better understand the MDP, let’s solve the Shortest Path Problem using the MDP approach:
Given the above representation, our goal here is to find the shortest path between ‘A’ and ‘D’. Each
edge has a number linked with it, this denotes the cost to traverse that edge. Now, the task at hand
is to traverse from point ‘A’ to ‘D’, with minimum possible cost.
In this problem,
• The set of states are denoted by nodes i.e. {A, B, C, D}
• The action is to traverse from one node to another {A -> B, C -> D}
• The reward is the cost represented by each edge
• The policy is the path taken to reach the destination
You start off at node A and take baby steps to your destination. Initially, only the next possible node
is visible to you, thus you randomly start off and then learn as you traverse through the network. The
main goal is to choose the path with the lowest cost.
Since this is a very simple problem, I will leave it for you to solve. Make sure you mention the answer
in the comment section.
The RL agent works based on the theory of reward maximization. This is exactly why the RL agent
must be trained in such a way that, he takes the best action so that the reward is maximum.
The collective rewards at a particular time with the respective action is written as:
The above equation is an ideal representation of rewards. Generally, things don’t work out like this
while summing up the cumulative rewards.
Let me explain this with a small game. In the figure you can see a fox, some meat and a tiger.
• Our RL agent is the fox and his end goal is to eat the maximum amount of meat before being
eaten by the tiger.
• Since this fox is a clever fellow, he eats the meat that is closer to him, rather than the meat
which is close to the tiger, because the closer he is to the tiger, the higher are his chances of
getting killed.
• As a result, the rewards near the tiger, even if they are bigger meat chunks, will be
discounted. This is done because of the uncertainty factor, that the tiger might kill the fox.
The next thing to understand is, how discounting of rewards work?
To do this, we define a discount rate called gamma. The value of gamma is between 0 and 1. The
smaller the gamma, the larger the discount and vice versa.
Explore Curriculum
Exploration, like the name suggests, is about exploring and capturing more information about an
environment. On the other hand, exploitation is about using the already known exploited information
to heighten the rewards.
• Consider the fox and tiger example, where the fox eats only the meat (small) chunks close to
him but he doesn’t eat the bigger meat chunks at the top, even though the bigger meat chunks
would get him more rewards.
• If the fox only focuses on the closest reward, he will never reach the big chunks of meat, this
is called exploitation.
• But if the fox decides to explore a bit, it can find the bigger reward i.e. the big chunk of meat.
This is exploration.
• Hyperparameters are variables that define the structure of the network. For example,
variables such as the learning rate, define how the network is trained.
• They are used to define the number of hidden layers that must be present in a network.
• More hidden units can increase the accuracy of the network, whereas a lesser number of
units may cause underfitting.
Grid Search
Grid search trains the network for every combination by using the two set of hyperparameters,
learning rate and the number of layers. Then evaluates the model by using Cross Validation
techniques.
Random Search
It randomly samples the search space and evaluates sets from a particular probability distribution.
For example, instead of checking all 10,000 samples, randomly selected 100 parameters can be
checked.
Bayesian Optimization
This includes fine-tuning the hyperparameters by enabling automated model tuning. The model
used for approximating the objective function is called surrogate model (Gaussian Process).
Bayesian Optimization uses Gaussian Process (GP) function to get posterior functions to make
predictions based on prior functions.
• Text Planning
• Sentence Planning
• Text Realization
Stemming algorithms work by cutting off the end or the beginning of the word, taking into account a
list of common prefixes and suffixes that can be found in an inflected word. This indiscriminate
cutting can be successful on some occasions, but not always.
Lemmatization, on the other hand, takes into consideration the morphological analysis of the words.
To do so, it is necessary to have detailed dictionaries which the algorithm can look through to link
the form back to its lemma.
Q12. Explain Fuzzy Logic architecture.
• Fuzzification Module − The system inputs are fed into the Fuzzifier, which transforms the
inputs into fuzzy sets.
• Knowledge Base − It stores analytic measures such as IF-THEN rules provided by experts.
• Inference Engine − It simulates the human reasoning process by making fuzzy inference on
the inputs and IF-THEN rules.
• Defuzzification Module − It transforms the fuzzy set obtained by the inference engine into
a crisp value.
• Knowledge Base
It contains domain-specific and high-quality knowledge.
• Inference Engine
It acquires and manipulates the knowledge from the knowledge base to arrive at a
particular solution.
• User Interface
The user interface provides interaction between the user and the Expert System itself.
Computer Vision is a field of Artificial Intelligence that is used to obtain information from images or
multi-dimensional data. Machine Learning algorithms such as K-means is used for Image
Segmentation, Support Vector Machine is used for Image Classification and so on.
Therefore Computer Vision makes use of AI technologies to solve complex problems such as Object
Detection, Image Processing, etc.
• In supervised classification, the images are manually fed and interpreted by the Machine
Learning expert to create feature classes.
• In unsupervised classification, the Machine Learning software creates feature classes based
on image pixel values.
Q16. Finite difference filters in image processing are very susceptible to noise. To
cope up with this, which method can you use so that there would be minimal
distortions by noise?
Image Smoothing is one of the best methods used for reducing noise by forcing pixels to be more
like their neighbors, this reduces any distortions caused by contrasts.
Q17. How is Game theory and AI related?
“In the context of artificial intelligence(AI) and deep learning systems, game theory is essential to
enable some of the key capabilities required in multi-agent environments in which different AI
programs need to interact or compete in order to accomplish a goal.”
Minimax is a recursive algorithm used to select an optimal move for a player assuming that the other
player is also playing optimally.
The following approach is taken for a Tic-Tac-Toe game using the Minimax algorithm:
Step 1: First, generate the entire game tree starting with the current position of the game all the way
up to the terminal states.
Step 2: Apply the utility function to get the utility values for all the terminal states.
Step 3: Determine the utilities of the higher nodes with the help of the utilities of the terminal nodes.
For instance, in the diagram below, we have the utilities for the terminal states written in the squares.
Let us calculate the utility for the left node(red) of the layer above the terminal:
Step 5: Eventually, all the backed-up values reach to the root of the tree. At that point, MAX has to
choose the highest value:
i.e. MAX{3,2} which is 3.
Therefore, the best opening move for MAX is the left node(or the red one).
To summarize,
Alpha-beta Pruning
If we apply alpha-beta pruning to a standard minimax algorithm, it returns the same move as the
standard one, but it removes all the nodes that are possibly not affecting the final decision.
In this case,
Minimax Decision = MAX{MIN{3,5,10}, MIN{2,a,b}, MIN{2,7,3}}
= MAX{3,c,2}
=3
Q3. Which algorithm does Facebook use for face verification and how does it
work?
Facebook uses DeepFace for face verification. It works on the face verification algorithm, structured
by Artificial Intelligence (AI) techniques using neural network models.
Input: Scan a wild form of photos with large complex data. This involves blurry images, images with
high intensity and contrast.
Process: In modern face recognition, the process completes in 4 raw steps:
Output: Final result is a face representation, which is derived from a 9-layer deep neural net
Training Data: More than 4 million facial images of more than 4000 people
Result: Facebook can detect whether the two images represent the same person or not
Q4. Explain the logic behind targeted marketing. How can Machine Learning help
with this?
Target Marketing involves breaking a market into segments & concentrating it on a few key
segments consisting of the customers whose needs and desires most closely match your product.
It is the key to attracting new business, increasing your sales, and growing the company.
The beauty of target marketing is that by aiming your marketing efforts at specific groups of
consumers it makes the promotion, pricing, and distribution of your products and/or services
easier and more cost-effective.
• Text Analytics Systems: The applications for text analytics ranges from search
applications, text classification, named entity recognition, to pattern search and replace
applications.
• Clustering: With applications including customer segmentation, fast search, and
visualization.
• Classification: Like decision trees and neural network classifiers, which can be used for text
classification in marketing.
• Recommender Systems: And association rules which can be used to analyze your
marketing data
• Market Basket Analysis: Market basket analysis explains the combinations of products
that frequently
co-occur in transactions.
Q5. How can AI be used in detecting fraud?
Data Extraction: At this stage data is either collected through a survey or web scraping is
performed. If you’re trying to detect credit card fraud, then information about the customer is
collected. This includes transactional, shopping, personal details, etc.
Data Cleaning: At this stage, the redundant data must be removed. Any inconsistencies or missing
values may lead to wrongful predictions, therefore such inconsistencies must be dealt with at this
step.
Data Exploration & Analysis: This is the most important step in AI. Here you study the relationship
between various predictor variables. For example, if a person has spent an unusual sum of money
on a particular day, the chances of a fraudulent occurrence are very high. Such patterns must be
detected and understood at this stage.
Building a Machine Learning model: There are many machine learning algorithms that can be
used for detecting fraud. One such example is Logistic Regression, which is a classification
algorithm. It can be used to classify events into 2 classes, namely, fraudulent and non-fraudulent.
Model Evaluation: Here, you basically test the efficiency of the machine learning model. If there is
any room for improvement, then parameter tuning is performed. This improves the accuracy of the
model.
Q6. A bank manager is given a data set containing records of 1000s of applicants
who have applied for a loan. How can AI help the manager understand which loans
he can approve? Explain.
This problem statement can be solved using the KNN algorithm, that will classify the applicant’s loan
request into two classes:
1. Approved
2. Disapproved
K Nearest Neighbour is a Supervised Learning algorithm that classifies a new data point into the
target class, depending on the features of its neighboring data points.
The following steps can be carried out to predict whether a loan must be approved or not:
Data Extraction: At this stage data is either collected through a survey or web scraping is
performed. Data about the customers must be collected. This includes their account balance, credit
amount, age, occupation, loan records, etc. By using this data, we can predict whether or not to
approve the loan of an applicant.
Data Cleaning: At this stage, the redundant variables must be removed. Some of these variables
are not essential in predicting the loan of an applicant, for example, variables such as Telephone,
Concurrent credits, etc. Such variables must be removed because they will only increase the
complexity of the Machine Learning model.
Data Exploration & Analysis: This is the most important step in AI. Here you study the relationship
between various predictor variables. For example, if a person has a history of unpaid loans, then
the chances are that he might not get approval on his loan applicant. Such patterns must be detected
and understood at this stage.
Building a Machine Learning model: There are n number of machine learning algorithms that can
be used for predicting whether an applicant loan request is approved or not. One such example is
the K-Nearest Neighbor, which is a classification and a regression algorithm. It will classify the
applicant’s loan request into two classes, namely, Approved and Disapproved.
Model Evaluation: Here, you basically test the efficiency of the machine learning model. If there is
any room for improvement, then parameter tuning is performed. This improves the accuracy of the
model.
Q7. You’ve won a 2-million-dollar worth lottery’ we all get such spam messages.
How can AI be used to detect and filter out such spam messages?
To understand spam detection, let’s take the example of Gmail. Gmail makes use of machine
learning to filter out such spam messages from our inbox. These spam filters are used to classify
emails into two classes, namely spam and non-spam emails.
Q8. Let’s say that you started an online shopping business and to grow your
business, you want to forecast the sales for the upcoming months. How would
you do this? Explain.
This can be done by studying the past data and building a model that shows how the sales have
varied over a period of time. Sales Forecasting is one of the most common applications of AI. Linear
Regression is one of the best Machine Learning algorithms used for forecasting sales.
When both sales and time have a linear relationship, it is best to use a simple linear regression
model.
Linear Regression is a method to predict dependent variable (Y) based on values of independent
variables (X). It can be used for the cases where we want to predict some continuous quantity.
Y=𝒃𝟎+𝒃𝟏 𝒙+ⅇ
Here,
• Y = Dependent variable
• 𝒃𝟎 = Y-Intercept
• 𝒃𝟏 = Slope of the line
• x = Independent variable
• e = Error
Therefore, by using the Linear Regression model, wherein Y-axis represents the sales and X-axis
denotes the time period, we can easily predict the sales for the upcoming months.
Q9. ‘Customers who bought this also bought this…’ we often see this when we
shop on Amazon. What is the logic behind recommendation engines?
E-commerce websites like Amazon make use of Machine Learning to recommend products to their
customers. The basic idea of this kind of recommendation comes from collaborative filtering.
Collaborative filtering is the process of comparing users with similar shopping behaviors in order to
recommend products to a new user with similar shopping behavior.
To better understand this, let’s look at an example. Let’s say a user A who is a sports enthusiast
bought, pizza, pasta, and a coke. Now a couple of weeks later, another user B who rides a bicycle
buys pizza and pasta. He does not buy the coke, but Amazon recommends a bottle of coke to user
B since his shopping behaviors and his lifestyle is quite similar to user A. This is how collaborative
filtering works.
Q10. What is market basket analysis and how can Artificial Intelligence be used
to perform this?
Market basket analysis explains the combinations of products that frequently co-occur in
transactions.
For example, if a person buys bread, there is a 40% chance that he might also buy butter. By
understanding such correlations between items, companies can grow their businesses by giving
relevant offers and discount codes on such items.
Market Basket Analysis is a well-known practice that is followed by almost every huge retailer in the
market. The logic behind this is Machine Learning algorithms such as Association Rule Mining and
Apriori algorithm:
• Association rule mining is a technique that shows how items are associated with each other.
• Apriori algorithm uses frequent itemsets to generate association rules. It is based on the
concept that a subset of a frequent itemset must also be a frequent itemset.
For example, the above rule suggests that, if a person buys item A then he will also buy item B. In
this manner the retailer can give a discount offer which states that on purchasing Item A and B,
there will be a 30% off on item C. Such rules are generated using Machine Learning. These are then
applied on items in order to increase sales and grow a business.
Q11. Place an agent in any one of the rooms (0,1,2,3,4) and the goal is to reach
outside the building (room 5). Can this be achieved through AI? If yes, explain how
it can be done.
This problem can be solved by using the Q-Learning algorithm, which is a reinforcement learning
algorithm used to solve reward based problems.
Let’s represent the rooms on a graph, each room as a node, and each door as a link, like so:
Now let’s try to understand how Q-Learning can be used to solve this problem. The terminology in
Q-Learning includes the terms state and action:
In the figure, a state is depicted as a node, while “action” is represented by the arrows. Suppose,
the Agent traverses from room 2 to room5, then the following path is taken:
Next, we can put the state diagram and the instant reward values into a reward table or a matrix R,
like so:
The next step is to add another matrix Q, representing the memory of what the agent has learned
through experience.
• The rows of matrix Q represent the current state of the agent
• columns represent the possible actions leading to the next state
Q(state, action) = R(state, action) + Gamma * Max [Q(next state, all actions)]
Here, Q(state, action) and R(state, action) represent the state and action in the Reward matrix R
and the Memory matrix Q.
Note: The Gamma parameter has a range of 0 to 1 (0 <= Gamma > 1).
• If Gamma is closer to zero, the agent will tend to consider only immediate rewards.
• If Gamma is closer to one, the agent will consider future rewards with greater weight
Finally, by following the below steps, the agent will reach room 5 by taking the most optimal path:
Q12. The crop yield in India is degrading because farmers are unable to detect
diseases in crops during the early stages. Can AI be used for disease detection in
crops? If yes, explain.
AI can be used to implement image processing and classification techniques for extraction and
classification of leaf diseases.
Image Segmentation: It is the process of partitioning a digital image into multiple segments so that
image analysis becomes easier. Segmentation is based on image features such as color, texture.
A popular Machine Learning method used for segmentation is the K-means clustering algorithm.
Feature Extraction: This is done to extract information that can be used to find the significance of
a given sample. The Haar Wavelet transform can be used for texture analysis and the computations
can be done by using Gray-Level Co-Occurrence Matrix.
Classification: Finally, Linear Support Vector Machine is used for classification of leaf disease.
SVM is a binary classifier which uses a hyperplane called the decision boundary between two
classes. This results in the formation of two classes:
1. Diseased leaves
2. Healthy leaves
Therefore, AI can be used in Computer Vision to classify and detect disease by studying and
processing images. This is one of the most profound applications of AI.