I hope this lesson helps you as well. Today, I am going to discuss workflows using examples that I found to be simple to understand.
Workflow of a machine learning project
This time, you have to pretend that you want to teach a computer to understand when someone says certain words, like “Alexa” or “Hey, Siri.”
First, you need to gather recordings of people saying “Alexa” and other words. You’ll also need recordings of various accents and voices to make sure the computer can recognize “Alexa” no matter who says it.
Next, you use a special program called a machine-learning algorithm to teach the computer how to recognize “Alexa” in the recordings. This algorithm learns by looking at lots of examples and figuring out what makes the word “Alexa” different from other words.
Once the computer has learned from the examples, you put its new skills to work. You might install the program on a device like a smart speaker, where it can listen for the word “Alexa” in real-life situations. But sometimes, the computer doesn’t get it right the first time. It might have trouble recognizing “Alexa” when spoken with different accents or in noisy environments.
So, you might need to go back to step one and collect more data, including recordings of people with different accents or in different environments. Then, you train the model again, hoping it gets better at recognizing “Alexa” each time. This process of collecting data, training the model, and improving it is called iteration.
The goal is to make the computer smarter over time, so it can understand “Alexa” no matter who says it or where they are. This same process applies to teaching computers to do other things. You collect data, train the model, and keep refining it until it works well in real-life situations.
Workflow of a data science project
Now let’s imagine you’re a teacher trying to improve student performance in a class. Here’s how data science might help:
You start by collecting data on your students, like their grades on assignments and tests, attendance records, and maybe even how often they participate in class discussions.
You look at all this data to see if there are any patterns or trends. For example, you might notice that students who attend class regularly tend to get better grades, or that certain topics are particularly challenging for the class as a whole.
Based on what you find, you come up with ideas to help your students do better. Maybe you decide to offer extra help sessions for topics that students struggle with, or you start giving more quizzes to encourage studying.
After trying out your ideas, you see if they make a difference in student performance. If more students start getting better grades, that’s great! If not, try something else. You keep using data to guide your decisions and improve your teaching methods.
So, whether you’re running an online shop, a coffee mug factory, or a classroom, data science can help you understand what’s going on, come up with ideas to make things better, and keep improving over time.
Working with an AI team
When working with an AI team for your project, clear goal-setting, providing relevant data, and realistic expectations are crucial for developing an effective AI solution to optimize growth.
Let’s see the example below:
You have a gardening project in mind, but you need assistance from an AI team to optimize plant growth. Here’s how you’d collaborate with them:
Project Initiation: You’ve got this exciting idea to use AI to help your garden thrive, but you’re not sure where to start. First, you’d need to define what success looks like. Let’s say you want to use AI to predict the optimal watering schedule for your plants to ensure they grow healthy.
Defining Success Criteria: To measure success, you’ll need data—information about your plants, soil conditions, weather patterns, and watering habits. This data will serve as the basis for the AI model. You might want to achieve a certain level of accuracy in predicting when to water your plants to avoid overwatering or underwatering.
Understanding Data Usage: The AI team will work with two sets of data: a training set and a test set. The training set consists of past data on plant growth and watering schedules, while the test set is used to evaluate the model’s predictions. The AI model learns from the training data and then tries to make accurate predictions based on the test data.
Data Size and Quality: The training data needs to be large enough for the AI model to learn effectively. It should include diverse information about different plants, soil types, and environmental conditions. However, even with a robust dataset, achieving 100% accuracy might not be feasible due to the complexity of gardening factors and potential data limitations.
Realistic Expectations: It’s essential to have realistic expectations about the AI model’s performance. Even if it doesn’t achieve perfect accuracy, it can still provide valuable insights and improve your gardening practices. Collaborating closely with the AI team to set achievable goals and understand the model’s capabilities is key to success.
KINOMOTO.MAG 