Unlock the Power of Machine Learning with Computer Science Café's Learning Resources

EMERGING TECHNOLOGIES
MACHINE LEARNING (IB)

HOME | IB | CASE STUDY 2023

MACHINE LEARNING

Machine learning is a subfield of artificial intelligence that focuses on the development of algorithms and statistical models that can enable a computer system to learn from and make predictions on data, without being explicitly programmed to perform the task. It involves feeding a large amount of data into a machine learning model, which then uses statistical analysis to identify patterns and relationships within the data. The machine learning model can then use these patterns and relationships to make predictions about new data, or to make decisions based on the input data.

There are three main types of machine learning:

Supervised learning: The system is trained on a labelled dataset, where the desired output is already known. The goal is to learn a mapping between the input and output variables.
Unsupervised learning: The system is given a dataset without any labels, and the goal is to identify patterns and relationships within the data. This is useful for exploratory data analysis and for finding structure in data.
Reinforcement learning: The system is trained through trial and error, receiving rewards or penalties based on its actions. The goal is to learn the best actions to take in a given situation to maximize the rewards.

Machine learning has numerous applications across various industries and domains, including image and speech recognition, natural language processing, and predictive analytics.

SUPERVISED LEARNING

Supervised learning is a type of machine learning where the system is trained on a labeled dataset, where the desired output is already known. The goal of supervised learning is to learn a mapping between the input variables and the output variables, so that the model can make predictions on new, unseen data.
In supervised learning, the training data consists of a set of input-output pairs, where the input is a set of features or attributes, and the output is the corresponding label or target. The model uses this training data to learn the relationship between the input and output, and to make predictions based on this relationship.
There are two main types of supervised learning algorithms: regression and classification.

Regression: The goal of regression is to predict a continuous output value based on the input features. For example, a regression algorithm might be used to predict the price of a house based on its location, size, and number of bedrooms.
Classification: The goal of classification is to assign a label or class to a given input based on the input features. For example, a classification algorithm might be used to predict whether an email is spam or not, based on its content and sender.

Supervised learning algorithms can be further divided into linear and non-linear algorithms, based on the type of relationship they model between the input and output variables. Linear algorithms make a simple linear assumption between the input and output, while non-linear algorithms can model more complex relationships.

UNSUPERVISED LEARNING

Unsupervised learning is a type of machine learning where the algorithm is trained on an un-labeled dataset, and the goal is to uncover hidden patterns or structures in the data. Unlike supervised learning, the algorithm does not have access to the correct output values, and it must find the structure in the data on its own.

In unsupervised learning, the algorithm is not given any specific target variable to predict. Instead, it must find structure in the data by grouping similar data points together, reducing the dimensionality of the data, or finding relationships between variables.

Examples of unsupervised learning algorithms include k-means clustering, principal component analysis (PCA), and deep neural networks. The choice of algorithm depends on the specific problem and the nature of the data.

Unsupervised learning has many real-world applications, such as market segmentation, anomaly detection, and data compression. It is also commonly used as a pre-processing step before using supervised learning algorithms, as it can help to reduce the dimensionality of the data and improve the accuracy of the predictions.

REINFORCEMENT LEARNING

Reinforcement learning is a type of machine learning where an agent learns to make decisions by interacting with an environment. The goal of the agent is to maximize a reward signal over time by taking actions in the environment.

In reinforcement learning, an agent receives a reward signal after taking an action in the environment. This reward signal provides information about the quality of the action taken and helps the agent to adjust its behavior. Over time, the agent learns to select actions that result in high rewards, and it becomes increasingly skilled at the task.

The process of reinforcement learning can be thought of as a trial-and-error process, where the agent takes actions, receives feedback in the form of rewards, and updates its behavior accordingly. This process continues until the agent has learned an optimal policy, which is a mapping from states to actions that maximizes the expected cumulative reward.

Reinforcement learning has many real-world applications, including game playing, robotics, and autonomous driving. It is used in situations where an agent must make decisions in an uncertain and changing environment and learn from its experiences to improve its performance over time.

MACHINE LEARNING WITH RECOMMENDATION SYSTEMS

Recommendation systems are a type of machine learning system that provide personalized recommendations to users. They use various algorithms to analyse large amounts of data about users, items, and interactions between them to make informed recommendations.

In a recommendation system, the main goal is to predict the preferences or ratings of users for items they have not yet interacted with. This is typically achieved by training a machine learning model on historical data about user-item interactions.

There are several types of machine learning algorithms that are commonly used in recommendation systems, including:

Collaborative filtering: This approach uses user-item interactions to find similar users and make recommendations based on their preferences. It can be implemented using memory-based or model-based techniques.
Content-based filtering: This approach uses the properties or features of items to make recommendations. It generates recommendations by finding items that are similar to items that the user has liked in the past.
Matrix factorization: This approach factorizes the large user-item interaction matrix into two smaller matrices, which can be used to make recommendations.

Once a machine learning model has been trained, it can be used to make recommendations to new users by predicting their preferences for items they have not yet interacted with. These predictions can be used to rank items and generate a list of personalized recommendations for each user.

Recommendation systems are widely used in various applications, such as e-commerce, music and video streaming, and social media. They play a crucial role in providing a personalized experience to users and improving user engagement.

COST FUNCTION

A cost function, also known as a loss function, is an essential concept in machine learning. It is a function that measures how well a machine learning model is performing by comparing its predicted output with the actual output.

The goal of a machine learning model is to make accurate predictions on unseen data. To achieve this, the model must be trained on a dataset of known inputs and outputs. During training, the model makes predictions on the inputs and the output is compared to the actual output to calculate the cost or loss.

The cost function is used to determine how well the model is performing by measuring the difference between the predicted output and the actual output. The cost is calculated as the average difference between the predicted output and the actual output over the entire training dataset.

The objective of machine learning is to minimize the cost or loss. This is done by adjusting the parameters of the model to make better predictions. The process of adjusting the parameters of the model to minimize the cost function is known as optimization.

There are different types of cost functions, depending on the type of problem being solved. For example, in a regression problem, the cost function may be the mean squared error, which measures the average squared difference between the predicted and actual values. In a classification problem, the cost function may be the cross-entropy loss, which measures the difference between the predicted probability distribution and the actual distribution of the classes.

A cost function is a measure of how well a machine learning model is performing by comparing its predicted output with the actual output. The goal is to minimize the cost or loss by adjusting the parameters of the model to make better predictions. Different types of cost functions are used depending on the type of problem being solved

F-MEASURE

The F-measure is a metric used to evaluate the performance of a machine learning model in binary classification tasks, where there are only two possible classes, such as "yes" or "no", "spam" or "not spam", etc.

In binary classification tasks, there are four possible outcomes of a prediction: true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN). A true positive is when the model correctly predicts the positive class, a false positive is when the model predicts the positive class but the true class is negative, a true negative is when the model correctly predicts the negative class, and a false negative is when the model predicts the negative class but the true class is positive.

The F-measure is a combination of precision and recall, two other commonly used metrics in machine learning. Precision is the ratio of true positives to the total number of positive predictions, and recall is the ratio of true positives to the total number of actual positive cases.

The F-measure is calculated as the harmonic mean of precision and recall. It is a single number that reflects both precision and recall, and provides a balanced view of the model's performance. The formula for the F-measure is:

F = 2 * (precision * recall) / (precision + recall)

The F-measure ranges from 0 to 1, with a higher value indicating better performance. A perfect F-measure of 1 indicates that the model has both high precision and high recall, meaning that it is correctly predicting all positive cases without any false positives or false negatives.

The F-measure is a metric used to evaluate the performance of a machine learning model in binary classification tasks. It is a combination of precision and recall, and provides a balanced view of the model's performance. A higher F-measure indicates better performance, with a perfect score of 1 indicating perfect precision and recall

HYPERPARAMETER

In machine learning, a hyperparameter is a setting or configuration value that is chosen before training a model, and it is not learned from the data like the parameters of the model. Hyperparameters affect how the model is trained and how well it performs, so choosing the right hyperparameters is an important part of developing a successful machine learning model.

Hyperparameters can be thought of as the knobs and switches that control the behavior of a machine learning model. For example, in a neural network, hyperparameters might include the number of hidden layers, the number of neurons in each layer, the learning rate, the batch size, the activation function, and the regularization parameter.

Choosing the right hyperparameters can be a challenging task, as different hyperparameter values can lead to very different performance of the model. Typically, hyperparameters are chosen through a process called hyperparameter tuning or hyperparameter optimization, which involves trying different hyperparameter values and evaluating their performance on a validation dataset.

Hyperparameter tuning can be done manually, by changing hyperparameters one at a time and evaluating the performance of the model, or it can be done automatically, using algorithms such as grid search, random search, or Bayesian optimization.

A hyperparameter is a setting or configuration value that is chosen before training a machine learning model and affects how the model is trained and how well it performs. Choosing the right hyperparameters is an important part of developing a successful model, and is typically done through a process of hyperparameter tuning or optimization.

TERMINOLOGY | CLASSIFICATION, REGRESSION, EUCLIDEAN DISTANCE AND COSINE SIMILARITY

Before looking a K-nn it would be useful for you to be familiar with the following terminology.

Classification: Classification is a type of supervised learning algorithm in which the goal is to predict the categorical class of a data point based on its features. The algorithm learns from a labelled dataset, in which each data point is associated with a known class label, and uses this information to make predictions on new, unseen data points.
Regression: Regression is another type of supervised learning algorithm in which the goal is to predict a continuous numeric value based on a set of input features. Regression algorithms learn from labeled datasets, where each data point is associated with a known numeric value, and use this information to make predictions on new data points.
Euclidean distance: Euclidean distance is a measure of the distance between two data points in a multi-dimensional space. It is calculated as the square root of the sum of the squared differences between the corresponding feature values of the two data points. Euclidean distance is commonly used as a distance metric in machine learning algorithms, such as k-nearest neighbours.
Cosine similarity: Cosine similarity is a measure of the similarity between two vectors in a multi-dimensional space. It measures the cosine of the angle between the two vectors and ranges from -1 (completely dissimilar) to 1 (completely similar). Cosine similarity is commonly used as a distance metric in text analysis and recommendation systems, among other applications.

Classification and regression are two types of supervised learning algorithms used in machine learning to predict categorical or continuous values, respectively. Euclidean distance and cosine similarity are distance metrics used to measure the distance or similarity between data points in a multi-dimensional space, commonly used in machine learning algorithms such as k-nearest neighbours and text analysis

K-NEAREST NEIGHBOUR (K-NN)

K-nearest neighbours, or k-NN for short, is a machine learning algorithm used in supervised learning tasks, such as classification and regression. It works by finding the k-nearest neighbours to a given data point based on a distance metric, and making predictions based on the labels of those neighbours.

In the context of recommendation systems, k-NN is used to suggest items to users based on the similarity of their preferences to those of other users. This is done by creating a dataset of users and their ratings for certain items, such as movies or books.

To make a recommendation, the algorithm first finds the k-nearest neighbours to a given user based on their ratings for the items. The distance between the user's ratings and those of other users is measured using a distance metric, such as Euclidean distance or cosine similarity. The k-nearest neighbours are the users who are most similar to the given user in terms of their preferences for the items.

Once the algorithm has identified the k-nearest neighbours, it recommends items to the given user based on the ratings of those neighbours. For example, if the neighbours have rated a certain movie highly, the algorithm might suggest that the given user watch that movie too.

K-NN is a simple but effective algorithm for recommendation systems, as it relies only on the similarity of users' ratings and does not require a model to be trained on the data. However, it has its limitations too. For example, it can be slow when working with large datasets, and it may not work well if there are not enough ratings for some items.

K-nearest neighbours is a machine learning algorithm used in supervised learning tasks, such as classification and regression. In recommendation systems, it is used to suggest items to users based on the similarity of their preferences to those of other users. The algorithm identifies the k-nearest neighbours to a given user and recommends items based on their ratings. It's a simple but effective algorithm for recommendation systems, but it has its limitations too.

SAMPLE DATA SHEET

ADDING A NEW ITEM
When a new item is added to the system, the k-nearest neighbours (k-NN) algorithm can be used to suggest it to users who are most likely to be interested in it based on their past ratings.

Here are the steps to implement this process:
1. Add the new item to the dataset: When a new item is added to the system, it needs to be added to the dataset along with its features and attributes. For example, if the system is recommending movies, the new movie would be added to the dataset along with information about its genre, actors, director, etc.
2. Calculate the similarity of the new item to existing items: Once the new item is added to the dataset, the similarity of the new item to existing items needs to be calculated using a distance metric, such as Euclidean distance or cosine similarity. The similarity score is calculated by comparing the features of the new item to those of existing items in the dataset.

SAMPLE CALCULATION
d(New Movie, Star Wars) = sqrt((4.2 - 4.2)^2 + (1000 - 2900)^2)
= sqrt((0)^2 + (-1900)^2)
= 1900 (Euclidean distance between the new movie and Star Wars: Episode IV - A New Hope)

SAMPLE DATA WITH NEW MOVIE AND EUCLIDEAN VALUE CALCULATED

3. Identify the k-nearest neighbours to the new item: The k-nearest neighbours to the new item are the existing items in the dataset that are most similar to it. The number of neighbors (k) is a hyperparameter that needs to be set based on the specific problem being solved.
4. Recommend the new item to users: Once the k-nearest neighbors to the new item have been identified, the algorithm recommends the new item to users who have rated those neighbors highly. For example, if the new movie is similar to a popular action movie, the algorithm might recommend the new movie to users who have rated that action movie highly.
5. Update the recommendation system: As users rate the new item, their ratings can be used to update the recommendation system and improve the accuracy of future recommendations.

When a new item is added to the system, the k-nearest neighbours (k-NN) algorithm can be used to suggest it to users who are most likely to be interested in it based on their past ratings. The algorithm calculates the similarity of the new item to existing items, identifies the k-nearest neighbours, and recommends the new item to users who have rated those neighbours highly. As users rate the new item, their ratings can be used to update the recommendation system and improve the accuracy of future recommendations.

Based on the lowest Euclidean distance, we can see that The Lord of the Rings: The Fellowship of the Ring is the movie that is most similar to the new movie. Therefore, we could recommend The Lord of the Rings: The Fellowship of the Ring to users who have rated the new movie highly.

Machine Learning VS Artificial Intelligence

1: Write a paragraph to summarise what Machine Learning is? Can you keep this paragraph under 250 words?
2: Write a paragraph to summarise what Artificial Intelligence is? Can you keep this paragraph under 250 words?
3: Write a paragraph to summarise what the key differences between Machine Learning and Artificial Intelligence? Can you keep this paragraph under 250 words?

What is the main difference between supervised and unsupervised learning?
Can you give an example of a real-world application of supervised learning?
What is the goal of an unsupervised learning algorithm?
What is the difference between collaborative filtering and content-based filtering in recommendation systems?
Can you explain the process of reinforcement learning?
What is the role of the reward signal in reinforcement learning?
How does a machine learning model make predictions in a recommendation system?
What is matrix factorization and how is it used in recommendation systems?
Can you give an example of an application of reinforcement learning?
How does the choice of algorithm affect the performance of a recommendation system?

SUGGESTIONS
We would love to hear from you

SUBSCRIBE
To enjoy more benefits

We hope you find this site useful. If you notice any errors or would like to contribute material then please contact us.