哈尔滨工业大学人工智能研究院有限公司关于征集人工智能场景应用赋能技术研发计划专项2023年度项目预申报通知
1. 项目申请主体应做到申请材料齐全、建设内容明确、评价指标清晰、经济与社会效益可期，项目主体部门具备按期开工和项目实施能力；
2. 项目团队成员诚信状况良好，无在惩戒执行期内的科研严重失信行为记录和相关社会领域信用“黑名单”记录。
3. 申报项目受理后，原则上不能更改项目负责人。
4. 项目负责人限申报一个项目。
哈尔滨工业大学人工智能研究院有限公司关于征集人工智能场景应用赋能技术研发计划专项2023年度项目预申报通知
1. 项目申请主体应做到申请材料齐全、建设内容明确、评价指标清晰、经济与社会效益可期，项目主体部门具备按期开工和项目实施能力；
2. 项目团队成员诚信状况良好，无在惩戒执行期内的科研严重失信行为记录和相关社会领域信用“黑名单”记录。
3. 申报项目受理后，原则上不能更改项目负责人。
4. 项目负责人限申报一个项目。
人工智能研究院学术报告 第2023
报告题目InnovatingandInterpretingNeuralNetworks
报告摘要Deeplearninghasrecentlyachievedhugesuccessinmanyapplications,includingnaturallanguageprocessing,computervisionandmore.Inthesecases,deeplearningcanoutperformorcompetewithhumans.Itiswidelyrecognizedthatmachinelearning,especiallydeeplearning,isaparadigmshiftinmanyfields.However,therearestillmanychallengesahead.Ononehand,overthepastyears,majoreffortshavebeendedicatedtoarchitectureinnovationsinthefieldofneuralnetworks,leadingtomanyadvancedmodels.Althoughdeeplearningisinspiredbythecomputationoftheneuralsystem,currentdeeplearningsystemsfallshortofreflectingneuraldiversity.Ontheotherhand,despitethefactthatdeeplearningperformsquitewellinpractice,itisdifficulttoexplainitsunderlyingmechanismandunderstanditsbehaviors.Thesuccessofdeeplearningisnotwellunderpinnedbytheeffectivetheory.Lackinginterpretabilityhasbecomeaprimaryobstacletothewidespreadtranslationandfurtherdevelopmentofdeeplearningtechniques.Inthisproject,weproposequadraticneuronstoaddresstheneuraldiversityproblemindeeplearning,whereinnerproducts(whicharelinearoperations)arereplacedwithquadraticcounterpartswhosenon-linearityenhancestheexpressiveabilityoftheneuron.Further,weproposesoftthresholdingtoreplaceReLUactivationforsignalprocessingtasks.Wewillevaluatetheirfeasibilityinpracticalcomputervisionproblemsaswellasmedicalimagingproblems,therebyenrichingmachinelearningarmory.Wewillalsodevelopinterpretationmethodsfortheinnerworkingofneuralnetworksandaccountabletheoriesforthesuccessofdeepnetworks.