Benchmark Datasets - free edition
What is the InfoFrames free edition?
InfoFrames free editions is a FREE library that you can use to experiment with well known data sets in a new format (see InfoFrames Benchmark Datasets). Discover how InfoFrames summaries can speed up your work.
What are InfoFrames Benchmark Datasets?
We provide a list of most popular benchmark datasets available in the form of InfoFrames summaries, so that you can experiment with AI/ML in no time.
Quick Start
1 # update pip
2 pip update
3 # Install the package
4 # into python virtual environment
5 pip install -U infoframes
6
7 python
8 >>>import infoframes as if_api
1 # update pip
2 pip update
3 # Install the package
4 # into python virtual environment
5 pip install -U infoframes
6
7 python
8 >>>import infoframes as if_api
1 # Run latest Docker version
2 docker run -v `pwd`/mydata:/infoframes/data infoframes/infoframes:latest
3
4 # Now connect to the container and run python
5 python
6 >>> import infoframes as if_api
Available datasets:
OriginalDataset | InfoFrame’s Dataset (*) | CompressionRatio (**) | Model (***) | Quality Difference (****) | Training Time Difference (5*) |
|---|---|---|---|---|---|
MNIST | Version 1 | 183 | 183XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,94 s |
MNIST | Version 2 | 183 | XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,76 s |
Fashion – MNIST | Version 1 | 120 | XGBoost | 85,4% -> 80,9% | 31,65 s -> 2,67 s |
Fashion – MNIST | Version 2 | 120 | XGBoost | 85,4% -> 80,9% | 32,31 s -> 3,09 s |
Original Dataset | Info Frame’sDataset ⓘOriginal dataset transformed info InfoFrames format | CompressionRatio ⓘOriginal dataset size divided by InfoFrames summaries size | Model ⓘModel used in the benchmark | QualityDifference ⓘAccuracy difference between models trained on the original dataset and on the InfoFrames summaries | Training TimeDifference ⓘTraining time difference between models trained on the original dataset and on the InfoFrames summaries |
|---|---|---|---|---|---|
MNIST | Version 1 | 183 | 183XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,94 s |
MNIST | Version 2 | 183 | XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,76 s |
Fashion – MNIST | Version 1 | 120 | XGBoost | 85,4% -> 80,9% | 31,65 s -> 2,67 s |
Fashion – MNIST | Version 2 | 120 | XGBoost | 85,4% -> 80,9% | 32,31 s -> 3,09 s |
MNIST | Version 1 | 183 | 183XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,94 s |
MNIST | Version 2 | 183 | XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,76 s |
Fashion – MNIST | Version 1 | 120 | XGBoost | 85,4% -> 80,9% | 31,65 s -> 2,67 s |
Fashion – MNIST | Version 2 | 120 | XGBoost | 85,4% -> 80,9% | 32,31 s -> 3,09 s |
* Original dataset transformed into InfoFrames format
** Original dataset size divided by InfoFrames summaries size
*** Model used in the benchmark
4* Accuracy difference between models trained on the original dataset and on the InfoFrames summaries
5* Training time difference between models trained on the original dataset and on the InfoFrames summaries
MNIST | Version 1 | 183 | 183XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,94 s |
MNIST | Version 2 | 183 | XGBoost | 93,7% -> 89,2% | 28,07 s -> 2,76 s |
Fashion – MNIST | Version 1 | 120 | XGBoost | 85,4% -> 80,9% | 31,65 s -> 2,67 s |
Fashion – MNIST | Version 2 | 120 | XGBoost | 85,4% -> 80,9% | 32,31 s -> 3,09 s |
Notebooks with examples on how to use InfoFrames summaries are now available on GitHub.
If you didn’t find a set that you are interested in, contact us at [email protected]
How to use InfoFrames edition
Mateusz Wnuk about If in Eng