Google With its artificial intelligence, it is capable of predicting which machine learning models will produce the best results. In a recent blog post by Google, the Google AI team of researchers called dış non-policy classification ”or OPC suggests what they say. This is considered as a classification problem.
The team says their approach works with image inputs and scales to tasks, including vision-based robotic understanding. Google's software engineer Alex Irpannon-policy empowerment learning provides artificial intelligence model training with a robot, but does not evaluate. Moreover, it is pointed out that basic accuracy assessment is often very inadequate in methods that require the evaluation of multiple models.
According to the proposed solution, the OPC addresses this problem by assuming that the tasks at hand have no or no coincidence of how situations change and that the agents succeed or fail at the end of the experimental trials. The dual nature of the second of the two assumptions allows the assignment of two classification labels (“effective için for success or“ disaster başarısızlık for failure).
The OPC also uses a Q-learning algorithm to predict the total total rewards of actions. Representatives choose actions with the highest projected rewards, and their performance is measured by how effective the chosen actions are. Subsequently, classification accuracy serves as a non-policy evaluation score.
The Google AI team trained machine learning policies in simulation using non-policy empowerment learning, and then evaluated them using non-policy points from previous real-world data. It has been reported that an OPC variant (SoftOPC) performs best in predicting the final success rate in a robot gripping task. In future studies, researchers aim to explore tasks with noisy and non-binary dynamics.