LSTM 智能导向的电子信息实践教改探索  

Exploration of LSTM intelligence-oriented electronic information practice education reform

作　　者：包建荣[1] 秦艺鹏 刘超[2] 李杰 姜斌[1] BAO Jianrong;QIN Yipeng;LIU Chao;LI Jie;JIANG Bin(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou 310018,China;College of Information Engineering,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学通信工程学院,浙江杭州310018 [2]杭州电子科技大学信息工程学院,浙江杭州310018

出　　处：《实验技术与管理》2024年第8期222-229,共8页Experimental Technology and Management

基　　金：浙江省自然科学基金项目(LZ24F010005);浙江省普通本科高校“十四五”教学改革项目(jg20220225);杭州电子科技大学2023年度实验技术专项项目(SYZD202301,SYYB202307);浙江省“十四五”研究生教学改革项目(浙学位办[2023]1号,序号172)。

摘　　要：针对传统电子信息专业教学中存在的问题,提出融合AI技术的针对性LSTM智能导向实践教学模型,构建了因材施教的人才培养体系。结合AI技术构建融合平台,包括结合深度学习技术在电子信息基础理论与实践课程中有针对性地推荐课程;借鉴强化学习思想和组内竞争思想助力学生竞赛创新;采用LSTM对学生个体进行建模,用以准确评估学生的自身能力,以便教师对他们进行因材施教。该教改方案实施以来,实验班相比对照班在获奖门类和数量上都有显著提高,验证了所提教改方案的有效性。[Objective]The combination of digital education and iterative upgrading of traditional education models has achieved good experimental results.Using digital empowerment to reform teaching models and intelligent overall management,diversified teaching models can be promoted and the efficiency of education operation and management can be fundamentally improved.In response to problems existing in the course design and teaching process of electronic information majors,this article builds an integrated teaching platform and proposes a quantitative student algorithm combined with long short-term memory(LSTM)technology to assist teachers in providing targeted training for students.[Methods]By modeling individual student models and using LSTM training,a student development model is constructed with individual students as the subject and parameters such as the assessment of students'motivation in class,records of correctness of practice problems,and awards for participation in competitions.Drawing on artificial intelligence conceptions,the intelligent prediction data process is used for practical education reform,and one-hot coding is used to encode individual student data.In addition,for the characteristics of individual student training data,an attention mechanism is introduced to the LSTM model,and a neural network is added to each state output of the original LSTM model.It makes the LSTM model selectively rely on all previous data records of the students instead of only relying on a single data input from the previous step when predicting the output of the students'data,which can make the prediction of the output results more accurate.In this paper,the LSTM network is combined with attention mechanisms to input all time steps of the LSTM hidden states and then input sequences into an attention model to compute the attention weights for each time step.Then,the computed attention weights are added to the input sequences to obtain a weighted input vector,which is fed into the LSTM network to perform the next prediction s

关 键 词：电子信息 人工智能 LSTM 创新人才培养 实践教学 

分 类 号：G642.0[文化科学—高等教育学]