Congratulations for reaching this far into the course. We've covered a lot about managing Machine Learning projects. Looking back, you learned about identifying the business value for Machine Learning, what it is, the phases within a project, using Machine Learning responsibly and ethically, discovering use cases and finally, the factors involved in managing Machine Learning projects. In this video, I'll review some of the core topics covered in each module. Module 1 was a brief course introduction. I provided an overview of the course objectives and topics we would and wouldn't cover. In Module 2, identifying the business value for using ML, you learned about some preliminary examples of ML solutions. This is where you uncovered how ML technology can be used to advance a company's mission no matter how big or small in size. You learned about the phases within a Machine Learning project. Once again, they are; assessing the problem, collecting and preparing the data. These are the input examples. Training the ML model, evaluating and validating model, and finally deploying it when it's ready. By now you also know that there are several considerations with the each phase, and that ML projects in particular don't follow a linear progression. It's common to go back a few steps and iterate before you can complete the project. For phase 1, you learned how to assess Machine Learning problems against a two-factor grid, from simple to impossible, and from specific to ambiguous. If you had to place your use case on these axes, do you remember where most ML problems typically fall? That's right. When assessing if a problem is suitable for ML, the problem should be challenging and clear. Module 3, defining ML as a practice provided you with the vocabulary and framework to work with ML. Specifically, you learned about the two main types of Machine Learning, supervised and unsupervised. Supervised learning problems fall into one of two categories. Can you remember what they are? They are regression. For example, predicting the price of a house, or classification such as predicting a category, label, or name. Unsupervised learning is fundamentally different because it doesn't involve predicting something specific. Unsupervised learning is about uncovering patterns in your data. Seeing if your data naturally falls into different groups or clusters. For the remainder of the course, we'll focused on supervised learning. Assuming a focus on supervised learning, we then talked about common elements that make up the definition of Machine Learning. ML is a way to use standard algorithms to analyze data, to derive predictive insights, and make repeated decisions. Remember, algorithm is another word for model. They both refer to a preset combination of mathematical functions that learn to understand a pool of data. Moving onto Module 4 in our recap, you learned that formulating an ML problem involves three parts. First, choosing the right objective. This is where you have to choose a problem for which you can get the input features and the corresponding labels. Second is choosing the input features. These are all the factors that affect the decision or objective. Third, getting the labels. You can use historical data if you have it, or you can build a labeling system to collect data, or you can use a labeling service. Each choice depends on the problem and the business. Next, as you learn about the importance of data in ML, we also highlighted the best-performing models are continuously trained with new data. That's because the world is always changing and only recent data will have a record of those changes. Lastly, in Module 4, you learned about evaluating and validating a trained model. A best practice is to hold a portion of that data to test the model. We might show the model somewhere around 80 percent of the data, and have it learned from that. We then test the model on the remaining 20 percent. We know the correct answers for all of our data, because it is already labeled. So we can evaluate how well the model will do when faced with the data it hasn't seen. Let's move onto Module 5, which was all about using ML responsibly and ethically. You learn that biases can appear at every point of a Machine Learning project, particularly where decisions are made and are then reflected in the data. Human bias occurs at the problem assessment stage when someone decides who the solution will benefit. It appears at the point of data collection when someone decides what kinds of data are accessible and which ones to include or omit. At the point of labeling, what human biases are annotators introducing into the data. It can occur at the model level 2, does a particular objective puts certain subgroups of people at a disadvantage compared to the group in aggregate? Those biases will appear in the output of a model and users will see the effect. You should always aim to remove these biases where you can. Ultimately, responsible AI is successful AI. Again, make sure that your ML model works for all your users and improve their quality of life. Technology is most powerful when everyone can use it and benefit from it. Module 6 was all about discovering ML use cases in day-to-day business, which provided you with some common ML use cases that apply to any industry. The most common ML use cases that any business can look for are replacing rule-based systems, automating business processes, and understanding unstructured data. The next step beyond these is to personalize the user's experience with ML. When you've mastered the fundamentals, you can start thinking about creative uses of ML such as using it to generate music or images. Finally, in Module 7, we covered the core considerations for successfully managing ML projects. Understanding the business value, formulating a data strategy, creating governance plans around ML, and pulling together an ML team. It would consist of expertise you currently have and ones you can gain by either buying, borrowing, or building the necessary skills. Finally, culture. Innovation is a mindset shift and it use organization why principals to help cultivate the right mindset to surface the ideas that bring the most value to the organization. That's it. Congratulations for completing managing Machine Learning Projects with Google Cloud. By now you should have the foundational knowledge to start incorporating ML into your business. Feel free to go back and review the lessons in this course at anytime to solidify your learning. If you enjoyed this course and want to learn more, check out our course catalog at cloud.google.com/training. Refer to the Machine Learning and artificial intelligence learning path for more details. If you're looking to add expertise to your team, you can do so in two ways. Google Cloud offers exceptional consulting services to help you further refine and assess the feasibility of your ML projects. You can also work with the Google Cloud partner. You can learn more at cloud.google.com/partners. With that, we've reached the end of this course. Thank you for your participation, and we look forward to seeing you again in another one of our courses.
