Stay organized with collections
Save and categorize content based on your preferences.
This section is a continuation of the previous exercise. Please continue to work
on your own problem or building a model to identify whether an email in Gmail is
"important."
You will be able to enter your response to the prompt into the textbox. You can
then print or store you answers as a .pdf using the button at the
bottom of the page.
[[["Easy to understand","easyToUnderstand","thumb-up"],["Solved my problem","solvedMyProblem","thumb-up"],["Other","otherUp","thumb-up"]],[["Missing the information I need","missingTheInformationINeed","thumb-down"],["Too complicated / too many steps","tooComplicatedTooManySteps","thumb-down"],["Out of date","outOfDate","thumb-down"],["Samples / code issue","samplesCodeIssue","thumb-down"],["Other","otherDown","thumb-down"]],["Last updated 2025-08-25 UTC."],[],[],null,["# Try It Yourself: Formulating\n\n\u003cbr /\u003e\n\nThis section is a continuation of the previous exercise. Please continue to work\non your own problem or building a model to identify whether an email in Gmail is\n\"important.\"\n\nYou will be able to enter your response to the prompt into the textbox. You can\nthen print or store you answers as a .pdf using the button at the\nbottom of the page.\n\nExercise 7a: Your Problem, Formulated as an ML Problem\n------------------------------------------------------\n\nWrite down or click on the checkbox for what you think is the best technical solution for your problem. Our problem is best framed as: \nBinary classification \nUnidimensional regression \nMulti-class single-label classification \nMulti-class multi-label classification \nMultidimensional regression \nClustering (unsupervised) \nGenerative AI \nOther\nwhich predicts or generates:\n\nIf your solution is best framed as a predictive ML one, see the following questions. If your solution is framed as a generative AI one, see [MakerSuite](https://makersuite.google.com)---a browser based IDE that lets you experiment with different prompts for quickly prototyping with generative language models.\n\nExercise 7b: Cast your Problem as a Simpler Problem\n---------------------------------------------------\n\nWhen first starting out, simpler problem formulations are easier to reason about and implement. Take your given problem and state it as a binary classification or a unidimensional regression problem (or both).\n\nExercise 8: Design your Data for the Model\n------------------------------------------\n\nWrite the data you want the predictive ML model to use to make predictions.\n\n| Tips for Success |\n|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| - One row constitutes one piece of data for which one prediction is made. - Only include information that is available at the moment the prediction is made. - Each input can be a scalar or 1D list of integer, float, or bytes (including strings). - If an input has a structure different from a scalar or 1D list, you may wish to consider whether that is the best representation for your data. For example: - If a cell represents two or more semantically different things in a 1D list, you may wish to split these into separate inputs. - If a cell represents a nested protocol buffer, you may wish to flatten out each field of the nested protocol buffer. - Exceptions: audio, image and video data, where a cell is a blob of bytes. |\n\nExercise 9: Where the Data Comes From\n-------------------------------------\n\nWrite down where each input comes from. Assess how much work it will be to develop a data pipeline to construct each column for a row.\n\n| Tips for Success |\n|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| When does the example output become available for training purposes? - If the example output is difficult to obtain, you might want to revisit Exercise 5 (Using the Output), and examine whether you can utilize a different output for your model. Make sure all your inputs (except the output) are available at serving time (when the prediction is made), in exactly the format you are writing down. \u003c!-- --\u003e - If it is difficult to obtain all your inputs at serving time in exactly the same format, you may want to revisit Exercise 8 (Design your data for the model) to reconsider inputs, or Exercise 5 to reconsider when serving can be made. |\n\nExercise 10: Easily Obtained Inputs\n-----------------------------------\n\nAmong the inputs you listed in Exercise 8, pick 1-3 inputs that are easy to obtain, and that you believe would produce a reasonable, initial outcome.\n\n| Tips for Success |\n|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|\n| - In Exercise 6, you listed a set of heuristics you could use. Which inputs would be useful for implementing these heuristics? - Consider the engineering cost to develop a data pipeline to prepare the inputs, and the expected benefit of having each input in the model. - Focus on inputs that can be obtained from a single system with a simple pipeline. Starting with the minimum possible infrastructure is advisable when first starting out. |\n\nPrint or save\n-------------\n\nClick the button below to either print or save your responses as a .pdf.\n\n[Print Page](#buttons)"]]