ML is a subfield of artificial intelligence (AI) in which computer systems learn patterns from data to perform specific tasks or make predictions on unseen data without being explicitly programmed. In 1959, Arthur Samuel defined ML as a “field of study that gives computers the ability to learn without being explicitly programmed to do so.” To give clarity to the definition given to us by Arthur Samuel, let us unpack it using a well-known use case of ML in the banking industry.

Imagine we work in the ML team of a Fortune 500 bank in the heart of London. We are saddled with the responsibility of automating the fraud detection process as the current manual process is too slow and costs the bank millions of pounds sterling, due to delays in the transaction processing time. Based on the preceding definition, we request historical data of previous transactions, containing both fraudulent and non-fraudulent transactions, after which we go through the ML life cycle (which we will cover shortly) and deploy our solution to prevent fraudulent activities.

In this example, we used historical data that provides us with the features (independent variables) needed to determine the outcome of the model, which is generally referred to as the target (dependent variable). In this scenario, the target is a fraudulent or non-fraudulent transaction, as shown in Figure 1.1.

Figure 1.1 – A flowchart showing the features and the target in our data

In the preceding scenario, we were able to train a model with historical data made up of features and the target to generate rules that are used to make predictions on unseen data. This is an example of what ML is all about – the ability to empower computers to make decisions without explicit programming. In classic programming, as shown in Figure 1.2, we feed in the data and some hardcoded rules – for example, the volume of a daily transaction to determine whether it is fraudulent or not. If a customer goes above this daily limit, the customer’s account gets flagged, and a human moderator will intervene to decide whether the transaction was fraudulent or not.

Figure 1.2 – A traditional programming approach

This approach will soon leave the bank overwhelmed with unhappy customers constantly complaining about delayed transactions, while fraudsters and money launderers will evade the system by simply limiting their transactions within the daily permissible limits defined by the bank. With every new attribute, we would need to update the rules. This approach quickly becomes impractical, as there will always be something new to update to make the system tick. Like a house of cards, the system will eventually fall apart, as such a complex problem with continuously varying attributes across millions of daily transactions may be near impossible to explicitly program.

Thankfully, we do not need to hardcode anything. We can use ML to build a model that can learn to identify patterns of fraudulent transactions from historical data, based on a set of input features in it. We train our model using labeled historical data of past transactions that contain both fraudulent and non-fraudulent transactions. This allows our model to develop rules based on the data, as shown in Figure 1.3, which can be applied in the future to detect fraudulent transactions.

Figure 1.3 – An ML approach

The rules generated by examining the data are used by the model to make new predictions to curb fraudulent transactions. This paradigm shift differs from traditional programming where applications are built using well-defined rules. In ML-powered applications, such as our fraud detection system, the model learns to recognize patterns and create rules from the training data; it then uses these rules to make predictions on new data to efficiently flag fraudulent transactions, as shown in Figure 1.4:

Figure 1.4 – An ML model uses rules to make predictions on unseen data

In the example we just examined, we can see from Figure 1.1 that our training data is usually structured in a tabular form, made up of numerical values such as transaction amount and frequency of transactions, as well as categorical variables such as location and payment type. In this type of data representation, we can easily identify both the features and the target. However, what about textual data from social media, images from our smartphones, video from streaming movies, and so on, as illustrated in Figure 1.5? How do we approach these types of problems where the data is unstructured? Thankfully, we have a solution, which is called deep learning.

Figure 1.5 – An illustration of structured and unstructured data types

Deep learning is a subset of ML that mimics the human brain by using complex hierarchical models that are composed of multiple processing layers. The buzz around deep learning lies around the state-of-the-art performance recorded by deep learning algorithms over the years across many real-world applications, such as object detection, image classification, and speech recognition, as deep learning algorithms can model complex relationships in data. In Sections 2 and 3, we will discuss deep learning in greater detail and see it in action for image and text applications, respectively. For now, let us probe further into the world of ML by looking at the types of ML algorithms.

In the last section, we looked at what ML is, and we examined a use case where we had labeled data. In this section, we will look at the four main types of ML approaches to give us a base understanding of what each type does, and where and how they can be applied. The four types of ML algorithms are as follows:

• Supervised learning
• Unsupervised learning
• Semi-supervised learning
• Reinforcement learning

Let’s examine the four types of ML methods, which will serve as a building block for later chapters.

Supervised learning

In supervised learning, an ML model is trained using data made of features and a target. This enables the model to learn the underlying relationship in the data. After training, the model can use its newfound knowledge to make predictions on unseen data. For example, say you want to buy a house. You would consider factors such as the location of the house, the number of rooms, the number of bathrooms, whether it has a garden, and the type of property; these factors you would consider as the features, while the price of the house is the target. Perhaps after the TensorFlow exam, you can roll up your sleeves, scrape some housing data, and train a model to predict house prices based on these features. You can use your house prediction model to compare prices on real estate websites to close a good deal for yourself.

There are two types of supervised learning – regression and classification. In a regression task, the label is a numeric value, just like the example we gave previously in which the target is the predicted price of the house. Conversely, in a classification task, the label is a category, just like the fraud example we discussed previously, in which the goal was to detect whether a transaction was fraudulent or not fraudulent. In a classification task, the model will learn to categorize the target as classes.

When dealing with a classification task made up of two classes (for example, fraudulent and non-fraudulent transactions), it is referred to as binary classification. When there are more than two categories (for example, the classification of different car brands), it is referred to as multi-class classification.

Figure 1.6: An example of multi-label classification, where the model identifies multiple subjects within an image

Multi-label classification is another type of classification, and it is used for image tagging in social media applications such as Facebook and Instagram. Unlike binary and multi-class classification tasks in which we have one target per instance, in multi-label classification our model identifies multiple targets for each instance, as shown in Figure 1.6, where our model identifies a girl, a dog, a boy, and a cat in the given photo.

Unsupervised learning

Unsupervised learning is the opposite of supervised learning. In this case, there is no labeled data. The model will have to figure things out by itself. Here, an unsupervised learning algorithm is provided with data, and we expect it to extract meaningful insights from the unlabeled data, by identifying patterns within the data without relying on predefined targets – for example, an image you work on for a large retail store that aims to segment its customers for marketing purposes. You are given data containing the demographics and spending habits of the store’s customers. By employing an unsupervised ML model, you successfully cluster customers with similar characteristics into distinct customer segments. The marketing team can now create a tailored campaign targeted at each of the customer segments identified by the model, potentially leading to a higher conversion rate from the campaign.

Semi-supervised learning

Semi-supervised learning is a combination of supervised and unsupervised learning. In this case, there is some data with labels (i.e., it has both features and a target) and the rest is without labels. In such scenarios, the unlabeled data is usually the dominant group. In this case, we can apply a combination of unsupervised and supervised learning methods to generate optimal results, especially when the cost implication and time required to manually label the data may not be practical. Here, the model takes advantage of the available label data to learn the underlying relationships, which it then applies to the unlabeled data.

Imagine you work for a big corporation that collects a large volume of documents that we need to classify and send to the appropriate department (i.e, finance, marketing, and sales) for effective document management, and only a small number of the documents are labeled. Here, we apply semi-supervised learning, train the model on the labeled document, and apply the learned patterns to classify the remaining unlabeled documents.

Reinforcement Learning

In reinforcement learning, unlike supervised learning, the model does not learn from training data. Instead, it learns from its interactions within an environment; it gets rewarded for making the right decisions and punished for making wrong choices. It is a trial-and-error learning approach in which the model learns from its past experience to make better decisions in the future. The model aims to maximize reward. Reinforcement learning is applied in self-driving cars, robotics, trading and finance, question answering, and text summarization, among other exciting use cases.

Now that we can clearly differentiate between the different types of ML approaches, we can look into what the core components of an ML life cycle are and what steps we should take, from the birth of a project to its application by end users. Let us take a look at the ML life cycle.

Before embarking on any ML project, we must take into account some key components that can determine whether our project will be successful or not. And this is important because as data professionals who want to build and implement successful ML projects, we need to understand how the ML life cycle works. The ML life cycle is a sensible framework to implement an ML project, as shown in Figure 1.7:

Figure 1.7 – The ML life cycle

Let’s look at each of these in detail.

The business case

Before unleashing state-of-the-art models on any problem, it is imperative you take time to sit with stakeholders to clearly understand the business objectives or the pain points to be resolved, as without clarity, the entire process will almost definitely fail. It is always important to keep in mind that the goal of the entire process is not to test a new breakthrough model you have been itching to try out but to solve a pain point, or create value for your company.

Once we understand the problem, we can categorize the problem as either a supervised or unsupervised learning task. This phase of an ML life cycle is all about asking the right questions. We need to sit with the concerned team to determine what the key metrics that would define the project as a success are. What resources are required in terms of budget, manpower, compute, and the project timeline? Do we have the domain understanding or do we need an expert’s input into defining and understanding the underlying factors and goals that will define the project’s success? These are some of the questions we should ask as data professionals before we embark on a project.

For the exam, we will need to understand the requirements of each question before we tackle them. We will discuss a lot more about the exam before we conclude this chapter.

Data gathering and understanding

When all the requirements are detailed, the next step is to collect the data required for the project. In this phase, we would first determine what type of data we will collect and where we will collect it from. Before we embark on anything, we need to ask ourselves whether the data is relevant – for example, if we collect historical car data from 1980, would we be able to predict the price of a car in 2022? Would data be made available by stakeholders, or would we be collecting it from a database, Internet of Things (IoT) devices, or via web scraping? Would there be any need for the collection of secondary data for the task at hand? Also, we would need to establish whether the data will be collected all at once or whether it will be a continuous process of data collection. Once we have collected the data needed for the project, we would then examine the data to get an understanding of it.

Next, we would examine the data to see whether the data collected is in the right format. For example, if you collect car sales data from multiple sources, one source may calculate a car’s mileage in kilometers per hour and another source could use miles per hour. Also, there could be missing values in some of the features, and we might also encounter duplicates, outliers, and irrelevant features in the data we collected. During this phase, we would carry out data exploration to gain insights into the data, and data preprocessing to handle various issues such as formatting problems, missing values, duplicates, removal of irrelevant features, and handling outliers, imbalanced data, and categorical features.

Modeling

Now that we have a good understanding of the business needs, we have decided on the type of ML problem that we will address, and we also have good-quality data after completing our preprocessing step. We will split our data into a training split and keep a small subset of the test to evaluate the model’s performance. We will train our model to understand the relationship between the features and the target variable using our training set. For example, we could train our fraud detection model on historical data provided by the bank and test it out with our hold out (test set) to evaluate our model’s performance before deploying it for use. We go through an iterative process of fine-tuning our model hyperparameters until we arrive at our optimal model.

Defining whether the modeling process is a success or not is tied to the business objective, since achieving a high accuracy of 90 percent would still leave room for a 10 percent error, which could be decisive in high-stake domains such as healthcare. Imagine you deploy a model for early-stage cancer detection with an accuracy of 90 percent, which means the model would likely fail once for every 10 people; in 100 tries, it could fail about 10 times, and it could misclassify someone with cancer as healthy. This could lead to the individual not only failing to seek medical advice but also to an untimely demise. Your company could get sued and the blame would fall in your lap. To avoid situations like this, we need to understand what metrics are important for our project and what we should be less strict with. It is also important to address factors such as class imbalance, model interpretability, and ethical implications.

There are various metrics that are used to evaluate a model, and the type of evaluation depends on the type of problem we will handle. We will discuss regression metrics in Chapter 3, Linear Regression with TensorFlow, and classification metrics in Chapter 4, Classification with TensorFlow,.

Error analysis

We are not ready for deployment yet. Remember the 10 percent data that could tank our project? We will address that here. We perform an error analysis to identify the misclassified labels to identify why the model missed them. Do we have enough representative samples of these misclassified labels in our training data? We would have to determine whether we need to collect more data to capture these cases where the model failed. Can we generate synthetic data to capture the misclassified labels? Or was the misclassified data down to the wrong labeling?

Wrongly labeled data can hamper the performance of a model, as it will learn incorrect relationships between the features and target, resulting in poor performance on unseen data, making the model unreliable and the entire process a waste of resources and time. Once we resolve these questions and ensure accurate labels, we need to retrain and reevaluate our model. These steps are continuous until the business objective is achieved, and then we can proceed to deploy our model.

Model deployment and monitoring

After resolving the issues identified in the error analysis step, we can now deploy our model to production. There are various methods of deployment available. We could deploy our model as a web service, on the cloud, or on edge devices. Model deployment can be challenging as well as exciting because the entire point of building and training a model is to allow end users to apply it to solve a pain point. Once we deploy our model, we also monitor the model to ensure that the overall objectives of the business are continually achieved, and even the best-performing models can begin to underperform over time due to concept drift and data drift. Hence, after deploying our model, we cannot retire to some island. We need to continuously monitor our model and retrain the model when needed in order to ensure it continues to perform optimally.

We have now gone through the full length of the ML life cycle at a high level. Of course, there is a lot more that we can talk about in greater depth, but this is out of the scope of this exam. Hence, we will now switch our focus to looking at a number of exciting use cases where ML can be applied.

Beyond the car price prediction and fraud detection use cases, let’s look at some exciting applications of ML here. Perhaps it will get you fired up for both the exams and inspire you to build something spectacular in your ML journey.

Healthcare

HearAngel uses AI to automatically prevent hearing loss for headphone users by tracking users’ exposure to unhealthy levels of sounds from the headphones. Insitro uses knowledge of ML and biology for drug discovery and development. Other use cases of ML in healthcare include smart record keeping, data collection, disease outbreak forecasting, personalized medicine, and disease identification.

The retail industry

ML engineers are revolutionizing the retail industry by deploying models to enhance customer experience and improve profitability. This is achieved by optimizing assortment planning, predicting customer behavior, the provision of virtual assistants, inventory optimization, tracking customers’ sentiments, price optimization, product recommendation, and customer segmentation, among other use cases. In the retail industry, ML engineers provide value by automating cumbersome manual processes.

The entertainment industry

The entertainment industry currently applies ML/AI in automatic script and lyric generation. Yes, currently there are movie script-writing ML models. One such movie is a short science fiction movie called Sunspring (https://arstechnica.com/gaming/2021/05/an-ai-wrote-this-movie-and-its-strangely-moving/). Also, ML/AI is used for automatic caption generation, augmented reality, game development, target marketing, sentiment analysis, movie recommendation, and sales forecasting, among others. So, if you plan to carve a niche in this industry as an ML engineer, there is a lot you can do, and surely you can also come up with some new ideas.

Education

Readrly utilizes deep learning techniques to create personalized children’s stories, enhancing the learning experience for young readers. By tailoring stories to each child’s interests and skill level, Readrly supports children’s reading development in a fun and engaging way.

Agriculture

There are numerous use cases of ML in agriculture. ML/AI can be used for price forecasting, disease detection, weather prediction, yield mapping, soil and crop health monitoring, and precision farming, among others.

Here, we covered some use cases of ML. However, the exciting part is that, as an ML/DL engineer, you will be able to apply your knowledge to any industry as long as data is available. And that is the awesomeness of being an ML engineer – there are no restrictions. We have covered a lot already in this chapter; however, we have one more section to go, and it is a very important one, as it centers on the exam itself. Let’s jump in.

The TensorFlow Developer Certificate exam was designed and developed by Google to assess data professionals’ expertise in model building and training deep learning models with TensorFlow. The exams enable data professionals to showcase their skills in solving real-world problems with ML/DL, as shown in Figure 1.8.

Figure 1.8 – The goal of the exam

Let’s dig deeper into why you should take this exam.

Why take the exam?

One of the most compelling reasons why you should take the TensorFlow Developer Certificate is that it can help you get a job. The global AI market is expected to grow exponentially, reaching $2 trillion by 2030 according to a report by Statista (https://www.statista.com/statistics/1365145/artificial-intelligence-market-size/#:~:text=The%20market%20for%20artificial%20intelligence,nearly%20two%20trillion%20U.S.%20dollars).

This rapid growth is being driven by continued advancements in areas such as autonomous vehicles, image recognition, and natural language processing, powering a new wave of applications across a wide range of industries. This growth is expected to create an increased demand for deep learning specialists who can build cutting-edge ML solutions.

In light of this development, recruiters and hiring managers are on the lookout for skilled candidates who can build deep learning models with TensorFlow, and this certificate can help you stand out from the crowd. To further accelerate your job-hunting, Google has put together TensorFlow Certificate Network, which is an online database of certified TensorFlow developers across the globe, as shown in Figure 1.9. Hiring managers can easily find suitable candidates to build their machine learning and deep learning solutions using a range of filters, such as location and years of experience, as well as verifying a candidate based on their name.

Figure 1.9 – A map display of TensorFlow-certified developers

In addition to helping you get a first job, the TensorFlow Developer Certificate can also help you advance your career. If you’re already working with TensorFlow, the certificate can help you demonstrate your expertise to your employer. This can lead to promotions and raises.

Now that we have looked at some of the reasons why you should take the exam, the next logical step is to look at what the exam is all about. Let us do that.

What is the exam all about?

If you’re planning on becoming a certified TensorFlow developer, there are a few things you’ll need to know. Here’s a quick rundown of what you need to know to ace the TensorFlow Developer Certificate exam:

• TensorFlow developer skills
• Building and training neural network models using TensorFlow 2.x
• Image classification
• Natural language processing (NLP)
• Time series, sequences, and predictions

You can find the complete exam details here: https://www.tensorflow.org/static/extras/cert/TF_Certificate_Candidate_Handbook.pdf. However, this book covers each section of the exam in detail to help ensure success. The exam costs $100, but there is an option to apply for a stipend that, if approved, will allow you to only pay half the exam fee. The stipend must be used within 90 days of receiving it and is only valid for one attempt. To apply for the stipend, you must provide information about yourself, why you need the stipend, and your portfolio projects with TensorFlow if you have any. You can find more information about how to access the TensorFlow Education Stipend using this link: https://www.tensorflow.org/static/extras/cert/TF_Education_Stipend.pdf.

We have now covered what and why. Now, let us look at how you can ace the exam.

How to ace the exam

If you’re looking to become a certified TensorFlow developer, there are a few things you should know. First, you should be proficient in Python. Second, you’ll need to have a strong understanding of ML concepts and be able to use TensorFlow to build and train deep learning models with TensorFlow. If you’re not already familiar with Python programming, then Modern Python Cookbook – Second Edition by Steven F Lott is a good place to start.

Here are some tips to help you ace the TensorFlow Developer Certificate exam:

• Review the course material: Before taking the exam, be sure to review the materials for every topic in the TensorFlow candidate handbook in detail. Pay special attention to building and training models, since the exam is hands-on.
• Model building: In addition to reviewing the course material, it’s also important to get some hands-on experience with TensorFlow. Experiment with building models, covering each section of the exam requirement. This book will get you started with the core fundamentals of ML and walk you through each section of the exam in a hands-on manner, ensuring that you can comfortably build and train all sorts of models with TensorFlow – from simple linear models to complex neural networks.
• Understand the exam format: The exam is unlike many other exams. It is a five-hour coding exam with questions covering each section of the exam that we outlined. You will be given a task and asked to write code to solve it in PyCharm. So, you will need to spend some time mastering how to build, train, and save models in PyCharm before the exams. The exam is an open-book one, so you are allowed to use any resource you want during the exam.
• Practice, practice, practice: One of the best ways to prepare for the exam is to practice solving questions. You will find lots of hands-on practice questions in every chapter of this book, along with code files in the book’s GitHub repository. Additionally, you will find lots of datasets on the TensorFlow website and Kaggle.

Once you’ve completed this book, you should be ready to take the TensorFlow Developer Certificate exam.

When to take the exam

Depending on your experience, you may need more or less time to prepare for the exam. If you are familiar with TensorFlow already, with hands-on model-building skills, you may take between three weeks to two months to prepare for the exam. However, if you are completely new to TensorFlow, it is advisable to look at around six months to thoroughly prepare for this exam, as stipulated on the exam website. However, these rules are not cast in stone. Everyone is different, so go at your own pace.

Exam tips

After you sign up for the exam, you can take it within a six-month period. It is okay to set a target day for your exams early enough. The exam takes place in PyCharm, so you will need to take a few days or weeks to get used to PyCharm (if you’re not familiar with it) before the exam day. Here is an excellent video tutorial by Jeff Henton to help you set up your PyCharm environment: https://www.youtube.com/watch?v=zRY5lx-So-c. Ensure you install the stipulated version of PyCharm. You can also learn more about setting up the exam environment using this link: https://www.tensorflow.org/static/extras/cert/Setting_Up_TF_Developer_Certificate_Exam.pdf.

Before the exam, it helps to have a clearly planned study routine in which you can cover the outlined syllabus for the exam. This book will help you on this journey, so you should pay attention to the next chapters, as we will start coding and tackling the core components, which make up the exam henceforth. On exam day, I would advise you to find a quiet, comfortable space to take this exam. Ensure you are well rested and not going in exhausted, as the exam is five hours long. Also, try out your PyCharm and internet connectivity. Don’t panic, and read the questions thoroughly to have a clear understanding of what is required of you. Start from questions 1 to 5. Since the questions get more difficult, it is better to get the easy ones out of the way quickly and tackle the more difficult ones afterward.

However, you should pace yourself correctly. Your saved model will be graded each time you submit it, and you are allowed to submit as many times as you want within the stipulated 5-hour time frame until you achieve an optimal result. You can also run your model in Colab if this enables you to work faster, especially if you have a model running in PyCharm. Colab provides you with free GPU access to train your model. The exam will be graded only in PyCharm, so bear this in mind. Ensure you save the model you train in Colab, and move it to the directory where you are stipulated to save the model for the exam.

If you need help, you can use Stack Overflow. You can also look through the code we use in this book, or any other material you use to prepare for your exam. However, if a question proves too difficult, move on to other questions. When you are done, you can return to the difficult one and work your way through it to avoid losing all your time on a difficult question. Also, you are allowed to submit multiple times, so work on improving your models until you attain optimal performance.

What to expect after the exam

The exam ends after five hours exactly, but you can submit it before that. After which, you will receive a congratulatory email if you have passed. After passing this exam, you will now be a member of the Google TensorFlow Developer community, opening more doors of opportunity for yourself. Assuming you pass (and I hope you do), you will get your certificate in about a week, which will look like Figure 1.10, and you will be added to the Google TensorFlow community in about two weeks. The certificate is valid for a period of three years.

Figure 1.10 – TensorFlow Developer Certificate

Now, you know the topics, the time frame, the cost, how to prepare, what to do on exam day, and what to expect after the exam. With this, we have come to the end of this chapter. We have covered a lot of theory in this chapter, which will serve as the basis for the work we will do together in the upcoming chapters.

This chapter provided an overview of ML, deep learning, and the types of ML approaches. It also covered the ML life cycle and various ML use cases across different domains. We looked at a high-level overview of the TensorFlow Developer Certificate, along with information on the components of the exam and how to prepare for it. At the end of this chapter, you should have a good foundational understanding of what ML is and its types. You should now be able to determine which problems are ML-based problems and those that require classic programming. You should also be able to unpack ML problems into different types and be familiar with what it takes to prepare for the TensorFlow Developer Certificate exam by Google.

In the next chapter, we will look at what TensorFlow is, set up our environment, and start coding our way to the end of this book.

Let’s test what we learned in this chapter:

1. What is ML?
2. What is deep learning?
3. What are the types of ML?
4. What are the steps in the ML life cycle?
5. What is the TensorFlow Developer Certificate about?
6. What are the core areas of the exam?

To learn more, you can check out the following resources:

• TensorFlow Developer Certificate overview: https://www.tensorflow.org/certificate
• Hands-On Machine Learning with scikit-learn and Scientific Python: Toolkits by Amr, T., Packt Publishing
• Deep Learning: Methods and Applications by Li Deng, and Dong Yu: https://doi.org/10.1561/2000000039
• Movie written by algorithm turns out to be hilarious and intense: https://arstechnica.com/gaming/2021/05/an-ai-wrote-this-movie-and-its-strangely-moving/.
• Python Machine Learning – Third Edition by Sebastian Raschka and Vahid Mirjalili