Don't forget to list down assumptions in ML? #5
Let's go for a hike!
Assumptions refers to the basic principles or conditions that are assumed to hold true about the data or a particular algorithm. These assumptions shape the way algorithms are designed and applied. Consider the example of Linear Regression.
Assumptions related to Linear Regression are:
There is a linear relationship between the independent variables (features) and the dependent variable (target).
The observations in the dataset are independent of each other.
The residuals (the differences between the actual and predicted values) are normally distributed
The variance of residuals is constant across all levels of the independent variables.
There is little or no multicollinearity among the independent variables.
There is no auto-correlation in the residuals.
Violation of any of these assumptions leads to biased or misleading results.
Let’s understand the importance of listing down assumptions with the help of an example.
Suppose, you are planning for a hike in a remote location. You study the map of the region before starting your hike. This is important to keep you aware of the potential pitfalls, weather challenges and other conditions on your journey. This will help you make necessary arrangements and detour strategies in advance. Without a map, you may get lost or make wrong turns.
Listing down assumptions in machine learning is similar to drawing a map before a journey. It helps us understand the ‘rules’ our model is following. Without listing assumptions, we may miss important factors, leading to flawed models or inaccurate results. It is like navigating without a map. We may get lost or take wrong turns. As explained earlier in the case of linear regression, if we don’t take care of the assumptions, it may lead to unreliable or biased results.
By listing assumptions, we set a clear path. It increases the chances of building reliable and effective machine learning models.
Now, having some idea about the importance of listing down assumptions in general, let’s understand its importance from business perspective:
Clarity and transparency
By listing down assumptions, business ensures that machine learning model development process align with their specific goals and requirements. Assumptions clearly states the factors influencing their plans and decisions.
Effective communication
Clear documentations of assumptions simplify the communication among stakeholders and machine learning practitioners. It clearly states the potential limitations and possible solutions, leading to informed decisions and expectation management. This ensures that everybody is on the same page and understand the basis of a particular decision.
Cost efficiency
Identifying assumptions beforehand helps in making informed decisions about resource allocation. It prevents the business from wasting resources, time and money into a model which produce misleading results due to violated assumptions.
Risk Management
Assumptions often carry inherent risks. Understanding these assumptions allow the business to assess the risks associated with machine learning solutions and take measures to mitigate those risks.
Conflicts Resolutions
Listing assumptions helps to resolve conflicts or disagreements in collaborative settings. It provides a starting point for discussion and allows stakeholders to identify potential areas of uncertainty or disagreements.
Flexibility
If assumptions are clearly stated, it becomes easier to revisit and revise them based on the new information or change in circumstances. This flexibility is important to adjust for unexpected developments or requirements.
Compliance and ethics
Some of the assumptions can be related to regulatory laws and compliances. Violation of these assumptions may lead to reputational and financial damage to the firm business. Identifying such potential pitfalls beforehand can really help.
Customer Trust
Clear documentation of assumptions indicates transparency. It demonstrates a business’s commitment to ethical and reliable use of machine learning which helps to build business’s reputation among its customers.
Learning and improvement
Reflecting on past assumptions and their outcomes can help future decision-making processes. By documenting assumptions and their outcomes, business create a learning opportunity that can lead to improved strategies and decision-making in the future.
What else would you like to add to the list?
In the next edition of this newsletter, I will talk about assumptions related to data, models, algorithm selection, evaluation metrics etc. in detail. Will bring some interesting analogies for you to understand them better. Stay tuned for the same.
Curious about a specific ML topic? Let me know in comments.
Also, please share your feedbacks and suggestions. That will help me keep going.
See you next Friday!
-Kavita
“Develop a passion for learning. If you do, you will never cease to grow.” — Anthony J. D’Angelo
P.S. Let’s grow our tribe. Know someone who is curious to dive into ML and AI? Share this newsletter with them and invite them to be a part of this exciting learning journey.