Скачать или смотреть Solving the NaN Issue: How to Properly Convert MLMultiArray to Swift [Double]

Discover the reason behind the conversion issue from `MLMultiArray` to Swift's `[Double]` and learn the correct approach to avoid `NaN` results in CoreML predictions.
---
This video is based on the question https://stackoverflow.com/q/62887013/ asked by the user 'jbm' ( https://stackoverflow.com/u/4321521/ ) and on the answer https://stackoverflow.com/a/62892294/ provided by the user 'Matthijs Hollemans' ( https://stackoverflow.com/u/7501629/ ) at 'Stack Overflow' website. Thanks to these great users and Stackexchange community for their contributions.

Visit these links for original content and any more details, such as alternate solutions, latest updates/developments on topic, comments, revision history etc. For example, the original title of the Question was: Converting MLMultiArray to [Double] causes NaN

Also, Content (except music) licensed under CC BY-SA https://meta.stackexchange.com/help/l...
The original Question post is licensed under the 'CC BY-SA 4.0' ( https://creativecommons.org/licenses/... ) license, and the original Answer post is licensed under the 'CC BY-SA 4.0' ( https://creativecommons.org/licenses/... ) license.

If anything seems off to you, please feel free to write me at vlogize [AT] gmail [DOT] com.
---
Solving the NaN Issue: How to Properly Convert MLMultiArray to Swift [Double]

When working with CoreML in Swift, specifically when trying to convert the output of a prediction() call from MLMultiArray to a Swift array type, developers may encounter perplexing NaN values. This guide delves into the conditions leading to this situation and provides a clear demonstration of the solution.

Understanding the Problem

While the conversion process seems straightforward, the transition from MLMultiArray to a Swift array can become muddled if the underlying data type is not correctly accounted for. Below is the original conversion function used for creating an array of doubles:

[[See Video to Reveal this Text or Code Snippet]]

However, users notice receiving arrays with NaN values instead of correct numerical data:

Sample Output

[[See Video to Reveal this Text or Code Snippet]]

The output clearly shows that something is amiss. Instead of valid double values, there are instances of NaN cluttering the array.

Analyzing the Cause of NaNs

The culprit here lies in the data type mismatch. The original data from MLMultiArray was represented as Float32. When the conversion function attempts to treat this data as Double, it leads to undefined behavior, resulting in NaN values.

Key Points to Consider:

MLMultiArray may contain Float32, Float64, or other types. Confirming the type is crucial for conversion.

Directly binding to Double without casting can lead to interpreting the memory incorrectly.

The Solution: Proper Conversion to Double

To achieve the intended conversion without encountering NaN, a new function that accurately reflects the underlying type is necessary. The final solution shared in the update illustrates this:

[[See Video to Reveal this Text or Code Snippet]]

Breakdown of the Solution:

Create a Result array: Start with an array of Double initialized to the size of o.count.

Iterate through the MLMultiArray: For each element, convert its value to Double using .doubleValue and store it in the result array.

Return the Final Array: Once all elements are processed, return the resulting array, which should now reflect accurate values without NaN.

Advantages of the New Approach:

The solution inherently respects the type of data within MLMultiArray.

It eliminates the memory binding issues and provides a clear path to conversion, avoiding garbage data and undefined values.

Conclusion

Converting an MLMultiArray output in CoreML to a Swift array can lead to tricky scenarios, particularly when the data types are not accounted for correctly. By recognizing that the original data might still be in Float32 and using the appropriate conversion method, developers can successfully convert their datasets to Double arrays without encountering NaN values. Remember to always check the underlying type of your data to ensure seamless conversions.



By following these guidelines, you can handle MLMultiArray conversions confidently and mitigate any confusion arising from unexpected outputs. Happy coding!