Bias in training data can originate from several sources, including: 1. Historical Bias: Pre-existing societal biases that get embedded in the data. 2. Collection Bias: Bias introduced during the data collection phase, such as through unrepresentative samples. 3. Labeling Bias: Bias that occurs when assigning labels or categories to data. Avoiding these biases involves rigorous methodologies for data collection and processing, continuous monitoring for bias, and employing diversity in the teams that collect and analyze data.
![](/sites/default/files/default_images/Women%20in%20Tech%20profile%20picture.png)
- Log in or register to contribute
Contribute to three or more articles across any domain to qualify for the Contributor badge. Please check back tomorrow for updates on your progress.