算法

Search documents
奥特佳(002239) - 奥特佳2024年度网上业绩说明会主要内容记录表
2025-05-09 10:10
证券代码:002239 证券简称:奥特佳 奥特佳新能源科技股份有限公司 投资者关系活动记录表 编号:2025-003 | | 特定对象调研 □分析师会议 | | --- | --- | | 投资者关系活动 | □媒体采访 ☑业绩说明会 | | 类别 | □新闻发布会 □路演活动 | | | □现场参观 | | | □其他 (请文字说明其他活动内容) | | 参与单位名称及 | 不特定的投资者、网友 | | 人员姓名 | | | 时间 | 2025 年 5 月 7 日 15 点 30 分至 16 点 30 分 | | 地点 | 在 全 景 路演 平台 以线 上 文 字交 流的 方式 开 展 ,网 址 : | | | https://rs.p5w.net/html/145990.shtml | | 上市公司接待人 | 董事长 王振坤先生、总经理 丁涛先生、财务总监 沈军女士、 | | 员姓名 | 董事会秘书 窦海涛先生、独立董事 胡振华先生 | | | 本公司于 2025 年 5 月 7 日 15 点 30 分至 16 点 30 分在全 | | | 景网路演平台举办了公司 2024 年度业绩网上说明会,与广大 ...
百邦科技(300736) - 300736百邦科技业绩说明会、路演活动信息20250509
2025-05-09 09:18
证券代码:300736 证券简称:百邦科技 北京百华悦邦科技股份有限公司 4.行业以后的发展前景怎样? 投资者关系活动记录表 2.公司之后的盈利有什么增长点? 答:感谢您的关注!公司业务的增长点聚焦于两大核心业务板 块:依托在手机售后服务领域多年的深耕积淀,公司已成功构建 起涵盖手机维修业务与联盟业务的服务体系。展望未来,公司将 通过持续推动产品服务创新迭代、加速联盟业务数字化转型、深 化基于大数据和算法的自动化体系建设,以及全面提升数字化营 销能力等关键举措强化核心竞争力,以此实现两大业务板块的规 模化发展与价值提升。 3.你们行业本期整体业绩怎么样?你们跟其他公司比如 何? 答:感谢您的关注!目前行业整体数据尚未公开披露。就公司 自身而言,持续推进产品结构升级,优化服务品类组合。同时, 从横向对比来看,公司已建立专业化的售后服务体系,具备成熟 的渠道网络和品牌认知度,在手机售后服务领域具备一定的市场 占有率。结合公司年度战略计划,公司将持续深耕细分市场,以 期在行业变革中进一步巩固差异化竞争力。我们将持续关注市场 动态,灵活调整经营策略,努力以更好的业绩回报投资者。 编号:2025-001 | | ☐特定 ...
汽车辅助驾驶 究竟如何感知?
Zhong Guo Zhi Liang Xin Wen Wang· 2025-05-09 08:42
汽车辅助驾驶已经逐渐成为现代汽车的重要组成部分。它不仅能够提升驾驶的安全性,还能为驾驶员提供更加人性化的驾驶体验。而这一切都离不开感 知技术的支持。那么,汽车辅助驾驶究竟是如何通过感知技术来实现其功能的呢? 感知方案都有哪些? 蝙蝠通过超声波听声点位,人类通过眼睛和耳朵来感知空间,而在汽车上,则通过各种传感器进行感知。目前市场主流的组合辅助驾驶主要是 V+R+L+U的感知组合,其中R代表毫米波雷达,V代表摄像头,L代表激光雷达,U代表超声波雷达,此外,车内的车载驾驶员监控(DMS)也算作一个摄 像头。 举例来说,以比亚迪的天神之眼C为例,其配备了5个毫米波雷达与12个摄像头,包括3颗800万前视摄像头、4颗300万环视摄像头、4颗300万侧视摄像 头、1颗300万后视摄像头,也就是5R12V感知方案。 V+R+L+U们看到的世界是什么样的? 以奔驰为例,其前视采用双目摄像头。双目摄像头通过两个摄像头同时感知前方物体,并利用视差算法计算物体距离,从而实现较为精准的测距功能。 BEV鸟瞰图 举例来说,通过多个摄像头采集图像后融合,系统将采集图像通过矫正后,统一输入到神经网络来提取特征,然后通过注意力机制的神经网络 ...
中科院软件所联合河北医大一院研发新技术:通过观看乒乓球比赛识别情绪障碍
Huan Qiu Wang Zi Xun· 2025-05-08 11:28
Core Insights - A breakthrough technology has been developed by the Institute of Software Research, Chinese Academy of Sciences, in collaboration with the First Hospital of Hebei Medical University, which efficiently identifies emotional disorders such as anxiety and depression through eye movement analysis during natural viewing of table tennis matches [1][2] - This technology offers a non-invasive and convenient new solution for mental health monitoring, with results published in the prestigious journal "Frontiers in Neurology" [1] Summary by Sections Technology Development - The research team utilized dynamic sports videos, specifically table tennis matches, as natural visual stimuli combined with virtual reality (VR) technology and machine learning algorithms to collect eye movement data without the need for active participation from subjects [1] - The study involved 25 participants, including 12 emotional disorder patients and 13 healthy controls, who watched table tennis and tennis match videos while their eye movement characteristics were recorded using the EyeKnow eye-tracking system [1] Research Findings - Significant testing and machine learning analysis revealed that 11 eye movement features from the table tennis videos effectively distinguished emotional disorders, with "Gaze Entropy" achieving an accuracy rate of 88% [2] - A decision tree model trained on all significant features reached an overall accuracy of 92% and an area under the ROC curve of 0.94, outperforming traditional assessment scales [2] - The tennis video showed weaker distinguishing effects due to lower familiarity among participants, but some indicators still demonstrated potential for adaptation to new content as a biological marker for emotional disorders [2] Future Applications - The core advantage of this technology lies in its "natural and unobtrusive" nature, requiring no active cooperation or professional guidance, allowing for detection simply through watching everyday videos [2] - Future integration into smart TVs and mobile devices is anticipated, enabling at-home mental health monitoring [2] - The research leader, Professor Tian Feng, emphasized that this method integrates disease screening into daily life, providing a low-cost and highly compliant long-term monitoring tool, particularly useful for tracking medication effects and relapse warnings [2]
吉宏股份通过港交所聆讯 乘跨境电商东风深耕亚洲市场
Zheng Quan Shi Bao Wang· 2025-05-08 06:48
Core Viewpoint - The ongoing trend of A-share companies listing in Hong Kong continues, with Xiamen Jihong Technology Co., Ltd. (referred to as "Jihong Co.") recently passing the hearing process for its IPO, highlighting its significant achievements in the cross-border social e-commerce sector [1][2]. Group 1: Company Overview - Jihong Co. operates as a cross-border social e-commerce enterprise, focusing on a "data-driven, technology-driven" model that enhances customer shopping experiences by significantly reducing the time needed to find products [1]. - The company utilizes a "goods find people" business model, primarily targeting overseas consumers with a diverse range of Chinese consumer goods, leveraging major social media platforms for advertising [1][2]. Group 2: Business Performance - In the first quarter of 2025, Jihong Co. reported a revenue of 1.477 billion yuan, representing a year-on-year growth of 11.55%, while net profit reached 59.16 million yuan, up 38.21% year-on-year [3]. - The company has established six brands, including SENADA BIKES and Veimia, and has fulfilled over 48 million orders for approximately 17 million consumers, with a signing rate between 84.9% and 88.4% [2]. Group 3: Market Position and Strategy - According to a report, Jihong Co. ranks second among Chinese B2C export e-commerce companies in terms of revenue generated from social media e-commerce in Asia, holding a market share of 1.3% [2]. - The company is expanding its business in the Middle East, with plans to establish packaging production bases and trade platforms in Ras Al Khaimah, UAE, and Salalah, Oman, which will help mitigate tariff uncertainties and support long-term growth in the region [3].
AI芯片股的下一个难关是……
Sou Hu Cai Jing· 2025-05-07 13:09
Group 1 - AMD reported a strong Q1 2025 performance with revenue of $7.438 billion, a year-over-year increase of 35.90%, exceeding market expectations of $7.13 billion [3] - The data center segment, which includes AI graphics and CPUs, generated $3.674 billion in revenue, up 57.21% year-over-year, with an operating margin of 23.37% [3] - AMD anticipates Q2 2025 revenue of approximately $7.4 billion, with a non-GAAP gross margin expected to be 43%, impacted by $800 million in inventory and related reserve costs due to new export controls [3] Group 2 - Supermicro's Q3 2025 revenue was $4.6 billion, a year-over-year increase of 19.48%, but significantly below market expectations of $5.42 billion [6] - Non-GAAP diluted earnings per share for Supermicro fell 53.03% year-over-year to $0.31, also below market expectations of $0.50 [6] - Supermicro lowered its full-year revenue guidance from $23.5 billion-$25 billion to $21.8 billion-$22.6 billion due to delayed customer deliveries and increased inventory reserves [6] Group 3 - The AI semiconductor industry faces challenges due to new tariffs and export controls, significantly impacting companies like Nvidia and AMD [8] - Nvidia's CEO indicated that being excluded from the Chinese AI market would result in substantial losses, as the Greater China region accounted for 39% of its revenue in FY2024 [8] - The uncertainty from U.S. policies and tariffs may lead clients to postpone expansion and technology upgrades, affecting the upstream supply chain, including companies like Supermicro [7][9]
托普云农:政策东风劲吹智慧农业赛道,多维布局抢占高标准农田建设新蓝海
Quan Jing Wang· 2025-05-07 12:50
Core Viewpoint - The company is positioned to benefit from national agricultural modernization policies and high-standard farmland construction initiatives, which present significant growth opportunities in the smart agriculture sector [1][2]. Group 1: Policy and Market Opportunities - The national strategy emphasizes agricultural modernization and aims to enhance comprehensive agricultural production capacity and ensure food security, creating numerous market opportunities for the company [1]. - By 2030, the goal is to establish 1.35 billion acres of high-standard farmland, with a vision to convert all eligible permanent basic farmland into high-standard farmland by 2035, indicating a substantial demand for smart agriculture technologies and products [1]. Group 2: Technological and Operational Advantages - The company has a strong advantage in technology research and development, focusing on cutting-edge technologies such as intelligent perception, machine vision, AI algorithms, and agricultural big data [2]. - It has developed an integrated data collection system to monitor agricultural conditions and provides decision support through proprietary algorithm models, showcasing its capability in project implementation [2]. - The company is committed to increasing R&D investment, optimizing existing products and services, and launching new technologies and products that align with market needs [2]. Group 3: Strategic Initiatives and Industry Leadership - To capitalize on policy-driven growth opportunities, the company is actively expanding its market presence and strengthening collaborations with government agencies, enterprises, and agricultural companies across various regions [2]. - The company aims to leverage its industry influence to participate in the formulation of relevant industry standards, thereby guiding the direction of industry development [2].
如何利用抖音算法?这5条AI指令,拆解品牌爆量密码
Sou Hu Cai Jing· 2025-05-07 10:12
Core Viewpoint - Understanding and mastering AI is crucial for businesses in a recommendation algorithm-driven e-commerce platform, as it directly influences brand visibility and sales performance [1][4]. Group 1: Algorithm Understanding - The recommendation algorithm on platforms like Douyin (TikTok) determines the flow of traffic for brands, making it essential for practitioners to comprehend its mechanics [1][6]. - Algorithms do not categorize content or audiences in traditional ways; instead, they utilize vast features to describe individuals or content through vectorization [3][4]. Group 2: Content Strategy - Emphasizing "content is king" and diversifying selling points is aimed at providing more features for the algorithm, which is critical for effective strategy execution [4][5]. - A clear and quantifiable representation system for commercial content teams is necessary to enhance algorithm recognition and improve brand visibility [7][8]. Group 3: Brand Analysis Method - The article suggests a method for analyzing brand performance by leveraging AI to dissect successful advertising materials, focusing on various attributes such as category, format, duration, and emotional tone [12][13]. - AI can assist in identifying key features in successful content, allowing brands to refine their strategies and enhance their representation systems [20][21]. Group 4: Storytelling and Consumer Engagement - Storytelling in marketing, particularly through relatable scenarios, can effectively engage target consumers and enhance brand connection [10][11]. - The article provides a framework for creating engaging content that resonates with young female consumers, emphasizing the importance of relatable interpersonal dynamics [28][29]. Group 5: Continuous Improvement - Brands should continuously refine their content strategies by repeating key features to strengthen model recognition and avoid irrelevant representations that could lead to misalignment with target audiences [22][23]. - The use of AI in content creation can lead to innovative approaches, allowing brands to explore new narrative possibilities that resonate with their audience [33][34].
DeepSeekGRM带来新的推理Scaling路径
HTSC· 2025-05-07 07:25
证券研究报告 计算机 DeepSeek GRM 带来新的推理 Scaling 路径 华泰研究 2025 年 5 月 06 日│中国内地 动态点评 近日,DeepSeek 团队的新论文《Inference-Time Scaling for Generalist Reward Modeling》发布,引入了一种自我原则点评调优(SPCT)的方法, 提出了通用奖励模型在推理阶段的新算法,验证了推理阶段的扩展策略在效 率与性能上的双重优势。同时基于此方法推出 DeepSeek GRM 模型,27B 的参数能跑出目前 R1 模型 671B 参数相当的性能。我们认为该方法带来了 推理阶段新的 Scaling 方法,或暗示了 DeepSeek R2 的技术路径,模型迭 代加速可期。 SPCT 通过生成的方法获得奖励信号,拓展了推理 Scaling 路径 目前,强化学习已广泛应用于 LLM 的大规模后训练阶段,现有的模型已经 证明采用合适的学习方法,可以实现推理的可拓展性。但强化学习过程中的 挑战在于在人工规则之外的一般领域难以获得 LLMs 的准确奖励信号,这决 定了强化学习的效果。DeepSeek 团队发现在奖励模型中采 ...
Beta Bionics Inc(BBNX) - 2025 Q1 - Earnings Call Transcript
2025-05-06 21:30
Beta Bionics (BBNX) Q1 2025 Earnings Call May 06, 2025 04:30 PM ET Company Participants Blake Beber - Head - Investor RelationsSean Saint - Board Member, President & CEOStephen Feider - CFOMike Kratky - Senior Managing Director - Medical Devices & TechnologyMathew Blackman - Managing DirectorTravis Steed - Managing Director - Equity ResearchJeffrey Cohen - MD - Equity ResearchKelly L. Close - Founder & President Conference Call Participants Matt O'Brien - AnalystJeffrey Johnson - Senior Research AnalystBroo ...