Blogs

Question number 5 is sort of boring ('nice job Zara discouraging me to read on' you may say! the answer to this question is where the rubber meets the road- who knows maybe I don't wan too much competition in the game?:))

Question 5: Can you find meaningful representations of the data?

The fuel (input) to your machine learning practice is good data. The algo runs on numbers but without good features you can't really know whether the algo is operating successfully or not. So it is critical to choose "informative, discriminating and effective features." Wikipedia

If you have heard of labels in machine learning and now wonder what the difference between a feature and a label is, you are not alone. Read the answer here.

The following examples from Wikipedia, walk you through different instances of machine learning features in practice:

1. In speech recognition, features for recognizing phonemes can include noise ratios, length of sounds, relative power, filter matches and many others.
2. In spam detection algorithms, features may include the presence or absence of certain email headers, the email structure, the language, the frequency of specific terms, the grammatical correctness of the text.
3. In computer vision, there are a large number of possible features, such as edges and objects.

As a manager you may not fully understand what really features are from the technical perspective. This is perfectly fine, as long as you know that they exist and play a crucial role in the success of your ML practice; you can discuss them and challenge your team on the quality of the features they choose for the algo.

Last but not least, keep on mind that you can't aim for perfection. The algo can be trained to identify better representations. So keep on mind that experimentation will constantly play an important role while you are managing your ML practice. Plan the cost, time, and the quality of the project without discounting the fact that your team will need to have enough room to experiment as they progress.

The next post will be the last post in this series. Don't miss it, it will be fun!