No. Training data is used to build the estimator model and then is immediately and permanently deleted. The same applies for prediction tasks: as soon as the results are computed, data provided as input is also permanently deleted.
Absolutely. Upon signing up, you immediately get 5,000 credits to get started, which you can use when you want. In addition, each month, we automatically refill your free credits to 2,500 (if your credit balance is higher than 2,500, you keep this balance).
No. Upon sign up, you receive your free credits and your API key, and you can start using JustML right away.
Each call to JustML is charged based on the compute time required by the algorithm. Calls are charged in credits, per second of usage. One credit corresponds to one second of compute time using a reference machine (2 cores at 2.5 GHz, 8 GB of RAM). 5,000 credits correspond to 1h23min of compute time using this reference machine.
Yes, all accounts – on both Free and Pay as you go plans – get the 5,000 initial free credits, plus the monthly refill to 2,500 free credits.
In the Pay as you go plan, you can purchase credits to pay for compute time that exceeds your free credits. Your free credits are always used before your paid credits. There is no subscription fee – you buy credits and use them when you want. Contrarily to free credits, paid credits never expire and can be cumulated.
Your API calls are blocked when you don’t have any credits over, or don’t dispose of enough credits for certain calls. You can purchase extra credits, or wait until your free credits are refilled in the beginning of the next month. If you’re on the Free plan and decide to buy credits, your account will be upgraded to the Pay as you go plan. Note that there is no subscription fee, and you choose when to buy credits, and how many of them.
5,000 credits corresponds roughly to 1 USD, as per the table shown in the next paragraph.
There are three credit packages that can be purchased, as shown in the following table.
|Number of credits||Price (USD)|
|500,000||95 (5% reduction)|
|1,000,000||180 (10% reduction)|
For the moment, contact us at email@example.com to buy credits.
Compute times varies with the dataset. Typically, for a dataset with 2,000 rows and 60 columns, automatically selecting and fitting a model takes 3 minutes, and performing a prediction takes 3 seconds. But again, it all depends on the dataset size and characteristics (number of categorical variables, need to encode labels, etc.), and on each problem needs. Model selection can be configured to take longer if one wants to test and compare a larger number of algorithms.
JustML services are hosted on Amazon Web Services (AWS) servers in the US.
For the moment, JustML tests and compares supervised machine learning models – classifiers or regressors – included in scikit-learn and xgboost. JustML also selects data and feature preprocessors in order to build a complete machine learning pipeline.
The selected machine learning pipeline is built and deployed on JustML computing infrastructure, so that it can be reused at any time to generate new predictions. Besides that, JustML outputs the details of the pipeline if you wish to rebuild it from scratch in your own environment. For the moment, we don’t output the algorithm as code but provide a textual description with the steps involved, and for each step, an indication of the name of the algorithm (for example, ‘sklearn.svm.SVC’) and the associated hyperparameters.
Contact us at firstname.lastname@example.org.