Reinforcement Learning

Online reinforcement learning API.

post
Reward

https://api.mlrequest.com/v1/rl/reward
Updates a reinforcement learning model with a single reward.
Request
Response
Request
Headers
MLREQ-API-KEY
required
string
Your API key
Body Parameters
model_name
required
string
The name of this model
session_id
required
string
The unique id of this user session
reward
required
number
The reward assigned to the model for performing this action. Smaller is bad, larger is good.
Response
200: OK
{
"message": "OK",
"remainingTransactions": 12345,
"total_active_models": 3,
"timestamp": "2019-10-05T02:33:23.972Z"
}

post
Predict

https://api.mlrequest.com/v1/rl/predict
Returns a single reinforcement learning model prediction.
Request
Response
Request
Headers
MLREQ-API-KEY
required
string
Your API key
Body Parameters
model_name
required
string
The name of this model
session_id
required
string
The unique id of this user session
features
required
object
A map of feature names and their values
epsilon
optional
number
The percentage of time the model will explore new options expressed as a number between 0 and 1
negative_reward
required
number
The negative reward to assign an example if a positive reward never occurs. This reward is assigned 1 hour after prediction.
action_count
required
integer
The total number of actions for this model
action_list
optional
array
A list of actions for the model to choose from
Response
200: OK
Returns a list of actions ordered by a score that estimates how well they will perform. The first action in the list is the highest scored, and therefore most-likely to obtain a good result.
{
"message": "RL prediction",
"predict_result": [2, 1, 3, 7, 5, 4, 9, 0, 6, 8],
"model_name": "my-model",
"remainingTransactions": 12345,
"total_active_models": 3,
"timestamp": "2019-10-05T02:33:23.972Z"
}