Welcome to the CS204A Artificial Intelligence Extension topic. In this topic, we will be extending the work you are doing in your elective course. By the end of this topic, you will have familiarised yourself with the following:
- Current topics in AI.
- The significance and impact of current projects in the field of computer vision.
- The use of Orange software in the context of AI projects.
For CS205, you are required to conduct a Literature Review and Research Proposal as part of your assessment, to assist you in developing this for the Artificial Intelligence stream, we recommend using Overleaf. Overleaf is an online collaborative writing platform designed specifically for creating and editing documents using LaTeX, a typesetting system commonly used for academic and scientific writing. It allows multiple authors to work together in real time on the same document, making it ideal for collaborative research papers, theses, reports, and other academic publications.
With Overleaf, users can create LaTeX documents using a user-friendly online editor, eliminating the need to install and manage LaTeX on their own computers. LaTeX is known for its high-quality typesetting, which produces professional-looking documents with precise formatting, mathematical equations, and symbols.
Key features of Overleaf include:
- Real-time Collaboration: Multiple authors can work together on the same document simultaneously, allowing for seamless teamwork and efficient collaboration.
- Version Control: Overleaf automatically saves document versions and tracks changes, enabling users to review the document's history and revert to previous versions if needed.
- Rich Text Editor: Overleaf provides a user-friendly rich text editor that simplifies the writing process, making it accessible to users without prior LaTeX experience.
- Reference Management: Overleaf integrates with BibTeX and BibLaTeX for managing references and citations, streamlining the process of citing sources within the document.
- Built-in Templates: Overleaf offers a variety of templates for different document types, such as research papers, theses, and presentations, saving users time in setting up the document structure and formatting.
- Math Mode and Equations: Overleaf supports a built-in math mode for typesetting complex mathematical equations and symbols accurately.
- Cloud-Based Access: Overleaf is a cloud-based platform, that allows users to access their documents from any device with an internet connection.
- Collaborator Management: Authors can invite collaborators to work on their documents and manage permissions for editing and commenting.
Overleaf is widely used in academia and research communities due to its ease of use, collaborative features, and LaTeX capabilities. It provides a convenient and efficient way for researchers, students, and authors to write, edit, and publish academic and scientific documents with professional formatting and accuracy.
Streamline your Literature Review with Overleaf
Overleaf offers the following to ensure you can streamline your literature review process.
| Feature | Description |
|---|---|
| Version Control and History | Overleaf automatically tracks document versions and changes, allowing authors to review the document's history and revert to previous versions if needed. This version control feature is particularly valuable when multiple contributors are involved, ensuring that everyone is working with the latest version of the literature review. |
| Easy Referencing and Citation Management | Overleaf integrates with popular reference management systems like BibTeX and BibLaTeX. Authors can easily import and manage references, making it simple to cite sources within the document. This feature ensures that references are accurately formatted and consistent throughout the literature review. |
| Automated Formatting | LaTeX, the typesetting system used by Overleaf, automates document formatting, ensuring a professional and consistent layout. Authors don't have to worry about manually adjusting headings, margins, or citations. LaTeX handles these aspects, allowing authors to focus on the content of the literature review. |
| Built-in Templates | Overleaf provides a variety of templates tailored to different document types, including research papers, theses, and literature reviews. These templates come pre-configured with appropriate section headings, citation styles, and formatting. Authors can select a suitable template, saving time and effort in setting up the document structure. |
| Cross-Referencing and Labeling | Overleaf makes it easy to cross-reference sections, figures, tables, and citations within the document. Authors can use labels and references to link content throughout the literature review, ensuring accuracy and consistency in cross-referencing. |
| Error-Free Typesetting | LaTeX's typesetting system ensures that the literature review is free from common formatting errors, such as inconsistent font sizes or spacing. This guarantees a polished and professional-looking document, essential for academic and research publications. |
| Spell-Checking and Autocomplete | Overleaf offers built-in spell-checking and autocomplete features, which help authors catch spelling errors and speed up the writing process. These features reduce the risk of typos and enhance the overall quality of the literature review. |
In summary, Overleaf streamlines the writing and collaboration process by providing real-time collaboration, version control, automated formatting, and easy reference management. These features enhance efficiency, promote collaborative writing, and ensure the literature review is well-organized, properly formatted, and accurately referenced. Overleaf's cloud-based accessibility and automated typesetting contribute to a seamless writing experience, allowing authors to focus on the content and research quality of the literature review.
Guide to Using Overleaf
Watch the following video and refer to the step-by-step guide below:
Using Overleaf effectively for writing the literature review section involves several steps and tips to ensure a smooth and efficient writing process. Here are some step-by-step instructions and tips:
- Create a New Project: Sign in to Overleaf or create a new account if you don't have one. Click on "New Project" and select a template suitable for your literature review, such as a research paper template or a thesis template.
- Organize Your Content: Before diving into writing, outline the structure of your literature review. Identify key sections, such as introduction, methodology, findings, analysis, and conclusion. Break down each section into subsections as needed. Having a clear outline will help you stay organized and focused while writing.
- Import References: If you have a list of references in BibTeX or BibLaTeX format, import them into your Overleaf project. You can do this by uploading a .bib file or manually adding references through the "Add Files" option. Organize your references into different categories based on themes or topics.
- Start Writing: Begin writing your literature review using LaTeX syntax. Follow the outline you created, and use appropriate sectioning commands such as "\section{}" and "\subsection{}" to structure your document. Use "\cite{}" to cite references and create a bibliography at the end of the document using "\bibliography{}" or "\printbibliography{}."
- Collaborate with Co-Authors: If you are collaborating with co-authors, invite them to join your Overleaf project. Share the link to your project, and they can edit the document simultaneously in real time. Use the commenting feature to provide feedback and suggestions to each other.
- Use LaTeX Commands for Special Formatting: Take advantage of LaTeX commands for special formatting needs, such as creating lists ("\begin{itemize}" and "\end{itemize}"), inserting figures ("\includegraphics{}"), or displaying equations ("\begin{equation}" and "\end{equation}"). Overleaf has a rich library of LaTeX commands that can enhance the appearance of your literature review.
- Automate Cross-Referencing: Use labels and references to cross-reference sections, figures, and tables within your document. For instance, label a section with "\label{sec:introduction}" and refer to it later using "\ref{sec:introduction}" for automatic numbering and updating.
- Review and Edit: Regularly review and edit your literature review as you progress. Overleaf provides a real-time preview of your document, so you can see how changes affect the final output. Take advantage of Overleaf's spell-checking feature to catch any typos or spelling errors.
- Version Control and History: Take advantage of Overleaf's version control to track changes and collaborate effectively. If you make significant changes or experiment with the content, consider creating a new version to maintain a record of your progress.
- Back-Up Your Work: Overleaf automatically saves your work, but it's always a good practice to regularly export and save a copy of your document locally as a backup.
- Ask for Feedback: Once you have a draft of your literature review, share it with colleagues or advisors for feedback. Overleaf makes it easy to share your document with others even if they don't have an Overleaf account.
- Finalize and Export: Once you are satisfied with your literature review, finalize it and export it to your desired format, such as PDF or Word. You can do this by clicking on the "Menu" button and selecting the appropriate export option.
Artificial Intelligence (AI) is revolutionizing the world in ways that were once only imaginable in science fiction. With its ability to analyze vast amounts of data, learn from patterns, and make intelligent decisions, AI is transforming various industries and aspects of our daily lives. From healthcare and finance to transportation and entertainment, AI is making a profound impact on how we live, work, and interact with technology.
The following are some examples of where AI is currently being used, you may wish to consider these for your literature review and area of research.
AI and climate change
The impact of AI extends beyond industries to societal challenges like climate change and sustainability. AI is being leveraged in climate modelling, renewable energy optimization, and environmental monitoring to tackle pressing global issues. It also plays a crucial role in disaster response and relief efforts, aiding in the allocation of resources and providing critical insights for emergency management. Let's take a look at some of these:
| Application | Description |
|---|---|
| Climate Modeling and Prediction | AI is employed in climate modelling and simulation, which helps scientists better understand climate patterns and predict future climate changes. By analyzing large datasets and complex interactions, AI-driven climate models can provide more accurate forecasts, aiding in climate adaptation and mitigation strategies. |
| Renewable Energy Optimization | AI algorithms are used to optimize renewable energy sources like solar and wind power. AI can forecast energy demand, adjust power output, and manage energy storage systems efficiently. This ensures that renewable energy resources are utilized optimally, reducing reliance on fossil fuels and lowering greenhouse gas emissions. |
| Smart Grids and Energy Management | AI-powered smart grids can intelligently manage energy distribution, balancing demand and supply in real time. This helps reduce energy wastage and improves the overall efficiency of the power grid. Smart grids can also integrate renewable energy sources seamlessly into the existing energy infrastructure. |
| Climate Change Monitoring | AI-enabled remote sensing and satellite technology are used to monitor and analyze environmental changes, such as deforestation, ice melting, and changes in vegetation cover. These observations help in tracking the impact of climate change and formulating appropriate conservation and restoration plans. |
| Climate Policy and Decision-Making | AI can assist policymakers in analyzing the potential outcomes of different climate change policies. By simulating scenarios and evaluating their impacts, AI helps inform better decision-making, leading to more effective and evidence-based climate policies. |
| Precision Agriculture | AI is used in precision agriculture to optimize resource use and reduce environmental impact. AI-driven agricultural technologies enable farmers to monitor soil conditions, irrigation, and pest control more efficiently, resulting in reduced water consumption, minimized use of pesticides, and enhanced crop yields. |
| Climate Change Adaptation | AI can analyze climate data to identify areas vulnerable to climate change impacts, such as extreme weather events and sea-level rise. This information aids in planning and implementing climate change adaptation strategies to protect communities and infrastructure. |
| Climate Finance and Carbon Trading | AI can assist in carbon market trading and monitoring emissions data to facilitate carbon trading and offset projects. By automating emission reporting and verification, AI streamlines carbon credit transactions, encouraging companies to reduce emissions and invest in sustainable practices. |
| Environmental Monitoring and Conservation | AI-powered technologies like drones and intelligent cameras are utilized for wildlife monitoring, forest protection, and marine conservation. AI can identify and track endangered species, detect illegal activities, and analyze environmental threats to support conservation efforts. |
AI's ability to process vast amounts of data, analyze patterns, and make informed predictions makes it a valuable tool in the fight against climate change. Its potential for continuous improvement and innovation holds promise for more effective and sustainable solutions to environmental challenges.
Watch the following video on one way in which AI is aiding climate change.
AI and quantum computing
Advances in quantum computing are poised to revolutionize the field of artificial intelligence (AI), ushering in a new era of computational power and capabilities. Quantum computing harnesses the principles of quantum mechanics to perform computations that are exponentially faster and more complex than classical computing. As quantum computers become more practical and scalable, their impact on AI is expected to be profound.
One of the key areas where quantum computing will transform AI is in solving complex optimization problems. AI algorithms often encounter optimization challenges, such as finding the most efficient routes, optimizing resource allocation, or tuning model parameters. Quantum computing's inherent ability to explore multiple possibilities simultaneously, thanks to quantum superposition and entanglement, promises to drastically accelerate optimization processes, leading to more efficient AI models and solutions.
Quantum computing can also vastly improve AI's ability to handle large datasets. Machine learning algorithms require significant computational resources for training and processing extensive data sets. Quantum computers' parallel processing capabilities can dramatically speed up these tasks, enabling AI systems to handle immense amounts of information and achieve more accurate and comprehensive results.
Moreover, quantum machine learning algorithms are being developed to enhance AI's learning and pattern recognition capabilities. Quantum algorithms can efficiently process quantum data representations, allowing AI systems to detect complex patterns and correlations that might be beyond the reach of classical algorithms. This could lead to groundbreaking advancements in natural language processing, image recognition, and other AI applications.
Another crucial aspect where quantum computing will impact AI is in cryptography and cybersecurity. Quantum computers have the potential to break current encryption methods, posing a security threat to classical computing systems. However, quantum cryptography methods are also being developed, which can provide unbreakable security and protect sensitive AI applications and data.
Despite these promising developments, quantum computing and its integration with AI are still in their early stages. Challenges such as error correction, scalability, and algorithm development need to be overcome to harness the full potential of quantum computing in AI applications. Nonetheless, the intersection of quantum computing and AI holds enormous promise, paving the way for innovations that could revolutionize industries, solve complex global challenges, and unlock new frontiers of knowledge and understanding. As research and advancements continue, the synergy between quantum computing and AI will undoubtedly shape the future of technology and redefine the possibilities of what AI can achieve.
Let's take a look at some of ways Quantum Computing will benefit AI.
| Impact | Description |
|---|---|
| Improved Optimization Algorithms | Quantum computing can solve complex optimization problems faster and more efficiently, leading to better AI predictions and decisions. |
| Quantum Machine Learning | Emerging field exploring the use of quantum computing to enhance machine learning algorithms and process quantum data. |
| Quantum Simulations for AI Research | Quantum computers can simulate quantum systems, leading to breakthroughs in quantum chemistry and optimization, influencing AI algorithms. |
| Enhanced Data Analysis | Quantum algorithms can accelerate data analysis tasks and improve the efficiency of AI applications that rely on data processing. |
| Quantum AI Algorithms for Quantum Data | Developing algorithms to process and analyze quantum data generated in quantum systems or experiments. |
| Quantum Machine Learning Hardware | Exploring quantum processors as a platform for implementing certain machine learning models more efficiently. |
| Privacy-Preserving Machine Learning | Using quantum cryptography and secure protocols to enhance privacy protection in machine learning tasks. |
| Early Stage of Development | Quantum computing is still in its early stages, and large-scale, fault-tolerant quantum computers are not yet available. |
Watch the following video on quantum computing in AI.
Other topics in AI
While we have covered some of the exciting topics of AI, there is a range of ways AI is and can be used to benefit industry and society.
- Explainable AI (XAI): Focusing on developing AI systems that can provide transparent and interpretable explanations for their decisions and predictions. This is essential for building trust and understanding in AI applications, especially in critical domains like healthcare and finance.
- Reinforcement Learning: Advancements in reinforcement learning algorithms and techniques, enable AI agents to learn and improve through trial and error. Reinforcement learning is being applied in various domains, including robotics, game-playing, and autonomous systems.
- AI in Finance: Applications of AI in the financial sector include fraud detection, risk assessment, algorithmic trading, credit scoring, and customer service.
- AI in Gaming: AI is extensively used in gaming for creating intelligent non-player characters (NPCs), procedural content generation, and optimizing game mechanics.
- AI in Cybersecurity: AI-powered solutions are used for threat detection, anomaly detection, and real-time response to cyber threats.
Computer vision, a transformative field of artificial intelligence (AI), is revolutionizing the way we perceive and interact with the world around us. It empowers machines to interpret, understand, and extract meaningful information from visual data, such as images and videos. From autonomous vehicles navigating busy streets to medical imaging aiding in early disease detection, computer vision applications are reshaping various industries and impacting our daily lives in remarkable ways.
As computer vision technologies advance and become more accessible, their potential to transform various industries and enhance human lives continues to grow. From enhancing healthcare outcomes to revolutionizing transportation and retail, computer vision's impact on the world is profound, unlocking new opportunities for innovation and societal progress.
Gesture recognition
Gesture recognition is a sub-field of computer vision developing systems to interpret human gestures. These include hand gestures and facial movements. Such systems have a wide range of uses:
- translating sign-language
- allowing a car GPS or entertainment system to be controlled by gestures, reducing driver distractions
- allowing systems to be accessed in a hygienic manner with no physical contact
- enhancing gaming systems
- allowing people with limited mobility to interact with a computer system.
Challenges
One of the challenges of gesture recognition is learning to distinguish a deliberate gesture from a normal movement. You do not want to wave at a friend and find your device has interpreted this as an instruction. On the other hand, a system that consistently fails to interpret deliberate gestures would be frustrating.
Another challenge is learning to interpret gestures successfully from a wide range of faces and hands. Human faces and hands come in a wide range of sizes and shapes. A system interpreting hand gestures may need to interpret input from the hands of children, people wearing gloves and people with fewer than five digits.
One-shot learning
Traditionally computer vision systems are trained using a large data set. The idea behind one-shot learning is to train a model to recognise patterns with a much smaller training set. The idea is that the model will learn to generalize from the limited examples provided.
One-shot learning reduces the time required to obtain and label a large data set. It can also be valuable where data is difficult to collect and label.
An example of one-shot learning is the development of automatic passport scanners and airport smart gates. These need to be able to interpret the data from a single photo and identify whether the person at the gate matches the passport photo.
Read the following article on one-shot learning.
https://encord.com/blog/one-shot-learning-guide/
Underwater computer vision
As the name suggests, this field of computer vision is concerned with analysing underwater images. Underwater images present different challenges from analysing land-based images. Colours display differently, particles of sand and small sea life interfere with images, and the water moves constantly, potentially blurring images. Specific algorithms are required to address these issues.
Underwater computer vision has many applications:
- monitoring marine habitats
- studying marine life
- facilitating the deployment of underwater submersibles
- shipwreck and archaeological site discovery and study.
Watch the following video on recolouring underwater images.
Orange is an open-source data mining tool that is used for data preprocessing, visualisation and analysis. It can also be used to develop machine learning models. Its user-friendly design means it can be used by novices lacking strong programming skills. However, it also integrates easily with Python, with Python scripts being imported into Orange, and Orange libraries being used in Python.
Orange can be downloaded from the website here: Orange Data Mining - Data Mining
Orange has a YouTube channel containing short, informative videos on the use of the app here: Orange Data Mining - YouTube
Features and widgets
All data interactions within Orange are performed by interconnected widgets. Some of main widget types are described below:
- Data widgets allow data to be imported from a range of sources including databases, spreadsheets and CSV files. New data can also be inserted directly into a data widget.
- Transform widgets allow raw data to be prepared before analysis. A single dataset can be sampled, creating subsets, or multiple datasets can be merged. Preprocessing is conducted at this point.
- Visualize widgets offer a range of tools for visualising data such as scatter plots, bar charts, boxplots, classification trees and more. The visualisation tools are interactive. For example, you can select points from a scatterplot and create a data subset.
- Model widgets providing a range of models including KNN, random forest, SVM and naive Bayes.
- Evaluate widgets include predictions with logistic regression and confusion matrix.
- Unsupervised widgets include PCA, K-means, Hierarchical Clustering, and DBSCAN.
AI use
In the AI course you learned the steps of supervised machine learning. These steps can be conducted using Orange.
- After being entered or imported, data can be preprocessed with a range of methods. For example, where there are missing values, the app can insert random values, or values averaged from existing data. Discretization of continuous values converted them into discrete values.
- Orange can help with feature selection and remove values that are considered less useful for analysis. PCA transformation can be used.
- Data can be divided into training and testing datasets.
- A wide range of models are available for training.
- Testing can be conducted on the test data. Performance statistics are provided to help determine accuracy and precision.
