What the Tech?

Deep Learning and AI Demystified

This article is the third in a series that aims to demystify data science , machine learning, deep learning, and artificial intelligence (AI) – while exploring how they are interconnected.

In this final article, we focus on deep learning and AI and how today’s computers mimic the structure of the human brain to perform natural language processing, image recognition, and more. We’ll also discuss how deep learning and AI work together to power data and predictive analytics.

What is deep learning?

Deep learning is a subset of the methods used in machine learning. It’s an algorithmic approach – based on Artificial Neural Networks (ANN) – for implementing machine learning that helps us better understand AI concepts.

As its name suggests, ANN is based on human brain structure and our understanding of how the brain processes data through interconnected neurons. More simplified than the brain, ANN has discrete layers and data flows through a defined number of connections and directions before producing output. Like other deep learning models, ANN can be trained with or without a teacher (known as supervised, unsupervised, or semi-supervised learning).

When using deep learning to conduct image analysis, for example, the image is broken down into a matrix of consumable data that serves as the input data for the first layer. Each neuron in this layer passes the data to a second layer, and so on, until it reaches the final layer where the resulting output is produced – which may suggest that the image is that of a dog.

For each image that the ANN consumes, neurons on each layer assign or adjust the weighting it places on its input. This indicates how correct or incorrect the output is relative to the task. The total of the weights determines the result.

Diagram showing input, hidden layers, and output of an Artificial neural network

The science behind ANN has been around for more than seven decades but it requires powerful computational resources to produce results from large amounts of data. As such, interest in deep learning waned until the development of computing systems advanced enough to support it. Nvidia, for instance, played a key role in advancing deep learning algorithms since ANN models were trained with Nvidia graphic processing units (GPUs).

Understanding the role of deep neural networks

ANN alone does not enable deep learning, instead deep learning consists of different ANN architectures known as Deep Neural Networks (DNN). These include Recurrent Neural Networks, Deep Feed Forward, Deep Botlzman Machine, and others. Each has a unique set of strengths and use cases including speech recognition, natural language processing, image analysis, and social network filtering – with results surpassing human expertise in some instances.

Deep learning vs. machine learning

As mentioned above, deep learning is a subset of machine learning. However, it brings new capabilities that advance machine learning to a new level.

Unlike traditional machine learning, deep learning doesn’t require you to research, develop, and use new features to improve the performance of machine learning algorithms – the neural network does the work for you by learning how to select the most critical features.

Furthermore, deep learning can deliver near-human performance, even surpass it. To achieve this, machine learning requires the guidance and reinforcement learning of a data scientist who would need to manually identify and create new features.

A downside of deep learning is that it requires longer training times since the quantity of training data can be massive. It can also be difficult to explain deep learning results since it’s hard to interpret the output of collective neurons. On the other hand, machine learning provides a clear set of rules to help explain the results.

Traditional Machine Learning Vs. Deep Learning Comparison

Deep learning vs. data science

Often confused as the same thing, deep learning and data science differ in that data science is an interdisciplinary field that uses multiple processes to collect, clean, analyze, and visualize data; then develop predictive or prescriptive models to solve a task or problem.

Deep learning and machine learning techniques are frequently used in data science to help assimilate big data, identify patterns and new features from that data, and create the predictive/prescriptive models used in data analytics.

Understanding AI

AI is the capability of a device or software agent to display human-like intelligence such as observing, learning, taking action, and solving a problem or task autonomously.

AI has inhabited the realm of science fiction for years where intelligent machines walk, talk, think, and act like humans. Could fiction become reality? Yes, and it’s already here. Virtual assistants like Siri and Alexa use sophisticated voice recognition to translate, interpret, and act on our requests or questions.

Self-driving vehicles are also powered by AI. Data from cameras integrated into the body of vehicles can help recognize objects, road conditions, and other environmental factors that enable them to safely drive autonomously.

In the world of E-commerce, AI is powering smarter platforms that can predict our purchasing interests based on past purchases and website behavioral habits.

In the home, Nest, a smart thermostat, can learn for our heating and cooling preferences and automatically adjust the temperature. AI is also helping to keep our floors clean. Robotic vacuum cleaners use lasers to map a floor plan for smart navigation and obstacle avoidance.

AI, machine learning, deep learning: bringing it all together

Although each has unique attributes, AI, machine learning, and deep learning combine to provide powerful data and predictive analytics capabilities.

AI automates any task that requires cognitive decision making and problem solving. Machine learning is a subset of AI, but not all AI is created using machine learning. Instead, machine learning is a set of algorithms and methodologies that enable AI to process, learn, and analyze big data. Machine learning also enables AI to perceive images, voice, video stream, and other environmental sources. With this input AI can solve a problem or conduct a task.

As such, AI is dependent on machine learning and deep learning techniques to process and learn from data in an autonomous way, identify patterns and features, and make decisions or classifications.

Driven by technological innovation, computational power, and improved understanding of how the human mind works, AI is experiencing exponential growth. It’s progress is unstoppable, and while some worry that it is a threat to life as we know it – particularly in the labor market – at EastBanc Technologies we will continue to contribute to a smarter, better, and safer AI for the future.