-
公开(公告)号：US11880765B2
公开(公告)日：2024-01-23
申请号：US17074054
申请日：2020-10-19
发明人： Pin-Yu Chen , Yada Zhu , Jinjun Xiong , Kumar Bhaskaran , Yunan Ye , Bo Li
IPC分类号： G06Q30/00 , G06N3/08 , G06F40/279 , G06Q40/06
CPC分类号： G06N3/08 , G06F40/279 , G06Q40/06
摘要： A processor training a reinforcement learning model can include receiving a first dataset representing an observable state in reinforcement learning to train a machine to perform an action. The processor receives a second dataset. Using the second dataset, the processor trains a machine learning classifier to make a prediction about an entity related to the action. The processor extracts an embedding from the trained machine learning classifier, and augments the observable state with the embedding to create an augmented state. Based on the augmented state, the processor trains a reinforcement learning model to learn a policy for performing the action, the policy including a mapping from state space to action space.
-
公开(公告)号：US20220121921A1
公开(公告)日：2022-04-21
申请号：US17074054
申请日：2020-10-19
发明人： Pin-Yu Chen , Yada Zhu , Jinjun Xiong , Kumar Bhaskaran , Yunan Ye , Bo Li
IPC分类号： G06N3/08 , G06Q40/06 , G06Q10/10 , G06F40/279
摘要： A processor training a reinforcement learning model can include receiving a first dataset representing an observable state in reinforcement learning to train a machine to perform an action. The processor receives a second dataset. Using the second dataset, the processor trains a machine learning classifier to make a prediction about an entity related to the action. The processor extracts an embedding from the trained machine learning classifier, and augments the observable state with the embedding to create an augmented state. Based on the augmented state, the processor trains a reinforcement learning model to learn a policy for performing the action, the policy including a mapping from state space to action space.
-
公开(公告)号：US10360004B2
公开(公告)日：2019-07-23
申请号：US15443038
申请日：2017-02-27
发明人： Si Bin Fan , Bo Li , Nai Jie Li , Jia Sen Wu , Zi Ying Xin , Xiao Zhen Zhu
摘要： A system, method and computer program product to refine an original complex CFG into a simpler CFG showing interesting paths and reducing interfering paths with dynamic input for the state of program. The method receives/encodes dynamic user input in the form of annotations which encodes user's special interests or knowledge of the program at run time, e.g., some assumptions of any variables appeared, which can be equations of variable and value or relationships between variables. The method then simplifies all the branching points in a generated AST (Abstract Syntax Tree) whenever possible by querying a SMT (Satisfiability Modulo Theories) solver with branching condition and the user annotations and by evaluating immediate values of expressions or eliminate unreachable parts in the CFG. Finally, the method generates a simplified CFG by simplified AST. This can assist a programmer to understand the code and facilitates correlating different basic blocks under a same scenario.
-
公开(公告)号：US20230214705A1
公开(公告)日：2023-07-06
申请号：US17566624
申请日：2021-12-30
申请人： International Business Machines Corporation , The Board of Trustees of the University of Illinois
发明人： Pin-Yu Chen , Nandhini Chandramoorthy , Karthik V Swaminathan , Jinjun Xiong , Devansh Paresh Shah , Bo Li
摘要： An input transformation function that transforms input data for a second machine learning system is learned using a first machine learning system, the learning being based on minimizing a summation of a task loss and a post-activation density loss. The input data is transformed using the learned input transformation function to alter the post-activation density to reduce an amount of energy consumed for an inferencing task and the inferencing task is carried out on the transformed input data using the second machine learning system.
-
-
-
搜索结果
4 条记录 (耗时 0.021 秒)
