In any typical application powered by machine learning,
whether that is a text classifier running in a web browser, or
a face detector running in a mobile phone, its machine learning code(or model) will be or close to 5% of the whole app.
Other 95%...👇
whether that is a text classifier running in a web browser, or
a face detector running in a mobile phone, its machine learning code(or model) will be or close to 5% of the whole app.
Other 95%...👇
Other 95% includes data, analysis, and software-related things.
A machine learning model being only 5% of the whole application often implies that we should be doing something else beyond tuning models.
A machine learning model being only 5% of the whole application often implies that we should be doing something else beyond tuning models.
Such as:
◆ Building irreproducible data preparation pipelines.
◆ Evaluating properly.
◆ Building irreproducible data preparation pipelines.
◆ Evaluating properly.
◆ Letting the error analysis guides data and model improvements.
◆ Keeping an eye on the important metrics to spot data/model drifts.
The image above was taken here: papers.nips.cc/paper/2015/fil…
More on what we should be doing: jeande.medium.com/ml-model-is-5-…
◆ Keeping an eye on the important metrics to spot data/model drifts.
The image above was taken here: papers.nips.cc/paper/2015/fil…
More on what we should be doing: jeande.medium.com/ml-model-is-5-…
• • •
Missing some Tweet in this thread? You can try to
force a refresh
