Auto ML | Coursera Community
Coursera Header
Solved

Auto ML

  • 16 January 2019
  • 2 replies
  • 115 views

Userlevel 3
Badge +4
How far has auto machine learning developed? Is it used in industry ? How effective they are ? Will there be impact on jobs because of auto ml .
icon

Best answer by Al French 16 January 2019, 19:35

the 3 most popular are automl autokeras and boxml and although they can all expedite the discovery of the "potentially" most useful hyperparameters none of them are really SOTA or fully production ready.
The main use case is to take the hyperparameters they provide and use them as the basis of your own further fine-tuning and tweaking before placing any models into production.
However constant research in this field will soon provide us with robust stable programs which will be production ready.
I personally use autokeras and boxml in tweaking reinforcement learning algos and find their results to be excellent starting points to work from.
View original

2 replies

Userlevel 1
Badge +1
the 3 most popular are automl autokeras and boxml and although they can all expedite the discovery of the "potentially" most useful hyperparameters none of them are really SOTA or fully production ready.
The main use case is to take the hyperparameters they provide and use them as the basis of your own further fine-tuning and tweaking before placing any models into production.
However constant research in this field will soon provide us with robust stable programs which will be production ready.
I personally use autokeras and boxml in tweaking reinforcement learning algos and find their results to be excellent starting points to work from.
Userlevel 1
Badge +1
This is also worth a look

# SMASH: One-Shot Model Architecture Search through HyperNetworks
An experimental technique for efficiently exploring neural architectures.

SMASH paper (https://arxiv.org/abs/1708.05344) and [video](https://www.youtube.com/watch?v=79tmPL9AL48).

SMASH bypasses the need for fully training candidate models by learning an auxiliary HyperNet to approximate model weights, allowing for rapid comparison of a wide range of network architectures at the cost of a single training run.

Reply

    Cookie policy

    We use cookies to enhance and personalize your experience. If you accept you agree to our full cookie policy. Learn more about our cookies.

    Accept cookies Cookie settings