本文首发于公众号"私募排排网"。 （点击↑↑ 上图查看详情 ） 半鞅私募基金简介 半鞅私募基金 成立于 2021年9月，核心投研团队为国内最早（2018）将机器学习运用于量化选股的团队之一。 半鞅私募基金 专注于挖掘高频Alpha，并自主开发了高频交易算法和高频交易系统。目前半鞅量化策略已有超过10年的经验，机器学习策略研 发经验超7年。投研框架成熟，且策略都已经历过多轮牛熊周期的考验。公司50%自有资金跟投在资管产品中，与投资者利益长期绑定。（点此 查看 半鞅私募基金旗下产品收益、核心团队及最新路演 ） 管理规模突破5亿:上线第二代 2023年 执行算法,提升短期趋势预 测;新增DMA、多空策略、中 波CTA等产品线 管理规模突破10亿，机器学习 2024年 领域取得显著突破，灰度上线 0 第三代执行算法,上线跨期套 管理规模突破20亿,Infra、 利模块,并开发新的展期策略 Alpha策略、量化CTA策略均取 2025年 得重大进展,包括上线新一代 · 流处理引擎、即将上线新一代 分布式计算平台、提高调仓频 率、提升交易收益占比等方 面。 秒 喜排排网 y 半额 户 核心团队 核心投研团队为国内最早(2 ...

转自：北京日报客户端 记者近日从京东获悉，今年该公司将面向2025届毕业生提供1.8万余个岗位。数据显示，截至4月30日， 京东体系员工总数已超过72万人，其中快递小哥、运输司机、分拣员工等一线员工总数超过50万人。 "非常惊喜！能在实习后通过转正述职，提前锁定正式校招offer（入职通知）。"去年正式入职京东的晓 韦说，公司为大学生人才设置了快速成长通道，他在入职后的短短一年间连获两次晋升，成长为一名能 够独当一面的采销人员。 京东集团雇主品牌负责人石玉介绍，公司在连续三年累计面向在校生提供5万多个岗位的基础上，今年 面向2025届毕业生再提供1.8万余个岗位，核心岗位薪资提升20%。同时，今年5月，京东启动了面向全 球技术人才招聘的"顶尖青年技术天才计划"，在新兴领域持续提供更多优质岗位，涵盖多模态大模型与 应用、机器学习、搜索推荐广告、空间与具身智能、高性能与云计算、大数据等前沿领域。 新技术催生新职业，公司近年来增添了许多新岗位，例如"大模型+"广告智能投放岗、"AI+"医疗服务 岗、家用机器人研发岗、无人机飞行师等等。 "有了'五险一金'，心里踏实也更有奔头。"今年3月成为京东外卖全职骑手的杨晶泽说 ...

Core Viewpoint - The article discusses the challenges faced by traditional and emerging companies in the NPU (Neural Processing Unit) market, emphasizing the need for a more integrated approach to matrix and general computing rather than relying on separate engines [1][4]. Group 1: Market Dynamics - The NPU IP licensing market is crowded with competitors offering various solutions, with many traditional CPU, DSP, and GPU IP providers entering the NPU accelerator space to maintain competitiveness [1][2]. - Leading IP companies have created similar AI subsystems that combine traditional cores with hardwired accelerators, resulting in a lack of differentiation in their offerings [2][4]. Group 2: Architectural Limitations - The existing architectures require algorithm partitioning to run on two engines, which works well for a limited number of algorithms but struggles with newer models like Transformers that require a broader set of graph operators [4][5]. - Traditional IP companies opted for short-term solutions by integrating matrix accelerators with existing processors, which has led to a technological trap as they now face the need for more advanced solutions [4][5]. Group 3: Long-term Challenges - The shift towards a programmable NPU capable of handling a wide range of graph operators is necessary but requires significant investment and time, which traditional companies have been reluctant to commit to [5]. - The "innovator's dilemma" is highlighted, where traditional companies must reconcile the need for new architectures with the legacy value of their existing IP cores, leading to a cycle of outdated solutions [5].

# SynBio团队 | 中科院苏州医工所马富强 在人工生命体精准编程的"黄金时代"，合成生物学作为融合工程学、计算机科学与分子生物学的交叉学科， 正通过"设计-构建-测试"的循环模式重塑生物制造范式。这一领域不仅被列为全球科技竞争的战略高地，更在 医药研发、碳中和、农业升级等关乎国计民生的赛道上展现出颠覆性潜力。 其核心突破点聚焦 三大支柱 ： 三大支柱技术如同精密咬合的齿轮，共同决定了细胞合成工厂的效率 。 【SynBioCon】 获 悉，近日， 苏州医工所马富强研究团队 围绕上述合成生物学三大支柱技术开展了系统工 作： 图2. 利用深度学习辅助生成新型启动子的两种方法 ： 左 图 ， 通过对现有启动子引入突变或随机生成新序列 来创建新的序列。这些新生成的序列通过启动子识别模型进行筛选，以验证其功能 ；右 图 ， 使用扩散模型 或生成对抗网络（GANs）来生成新的启动子。扩散模型通过添加高斯噪声逐步生成启动子序列。GANs 由生 成器和判别器组成，生成器负责生成假样本，而判别器用于区分真实样本和生成的假样本。通过训练过程不 断优化生成器的性能，使其能够生成更逼真的启动子序列。 工作1 ： 新型酶资源 的 ...

Core Viewpoint - The automotive industry is facing increasing challenges in ensuring the reliability and quality of integrated circuits and systems, particularly as vehicles become more reliant on advanced driver-assistance systems (ADAS) and software-defined functionalities [2][4][19]. Group 1: Challenges in Automotive Chip Development - The traditional development cycle for automotive chips is five to seven years, but the shift towards ADAS and complex infotainment systems has accelerated this process [2][4]. - Achieving automotive-grade quality with a defect rate below one part per million (DPPM) is a significant challenge, necessitating innovative testing methods [2][4]. - Manufacturers are under pressure to maintain low testing costs while ensuring high quality, creating a delicate balance [2][4][5]. Group 2: Advances in ADAS and Software-Defined Vehicles - ADAS has driven the automotive industry towards smaller technology nodes and more complex systems, transitioning to fully software-defined vehicles (SDVs) [4][5]. - The shift to advanced nodes below 5nm presents reliability and safety challenges, particularly for systems expected to operate for extended periods [4][5][19]. - Most new vehicles are currently at Level 2 or Level 3 automation, with increasing safety standards required for higher levels of automation [7][8]. Group 3: Testing and Quality Assurance - Automotive chips must undergo rigorous testing at three temperature extremes to simulate operational conditions, as defined by AEC-Q100 standards [9]. - Machine learning-based anomaly detection methods are increasingly used to enhance quality levels close to zero DPPM [9][10]. - Advanced fault models are being developed to better simulate common defects in silicon, improving testing accuracy [10]. Group 4: Virtual Testing and Predictive Maintenance - Virtual testing is becoming essential to reduce the complexity of real-world testing, allowing for parallel development and faster time-to-market [8][19]. - Continuous monitoring and feedback throughout the vehicle's lifecycle are critical, especially as more advanced nodes are introduced [19]. - On-chip monitoring and machine learning are being utilized to track performance degradation and predict failures [18][19]. Group 5: Future Directions in Automotive Testing - The industry is moving towards chiplet-based designs to improve yield and reuse rates while managing the complexity of advanced packaging [12][13]. - Acoustic and optical technologies are being employed to analyze inter-chip bonding characteristics, which are crucial for reliability [14]. - System-level testing is becoming a standard requirement to ensure that both hardware and software meet functional and non-functional requirements [16].

Core Viewpoint - The article emphasizes the integration of AI and multi-omics analysis, highlighting the importance of machine learning in biological data analysis and the educational offerings to enhance skills in this area [1][2][3]. Group 1: Course Features - The course is designed for beginners, providing a comprehensive introduction to R programming for bioinformatics analysis [1]. - It covers various popular directions in multi-omics research, including metabolomics, proteomics, microbiomics, and transcriptomics, keeping pace with scientific advancements [1]. - The teaching model includes one-on-one guidance and a flexible learning pace with live classes and recorded sessions [3]. Group 2: Course Content Overview - The first session focuses on interpreting CNS papers using Deepseek for efficient reading and summarizing multi-omics data analysis methods [2]. - Subsequent sessions cover the design of multi-omics research projects, programming basics in R, and machine learning applications in metabolomics, proteomics, and microbiomics [2][4][6]. - Advanced topics include the use of various machine learning models like xgboost, lasso, and random forests for intelligent data analysis [3][10]. Group 3: Practical Applications - The course includes hands-on experience with real CNS article source codes, allowing participants to replicate high-level research [3]. - It emphasizes the application of machine learning techniques in analyzing metabolomics and proteomics data, including regression models and network analysis [4][9]. - The integration of multi-omics data for comprehensive analysis is highlighted, showcasing the potential for significant insights in biomedical research [12][14].

Summary of SPS Commerce (SPSC) FY Conference Call - June 10, 2025 Company Overview - SPS Commerce operates a cloud-based network connecting retailers and suppliers for efficient supply chain information exchange, primarily focused on the ordering process [4][5] - The company has the largest EDI (Electronic Data Interchange) network in North America, emphasizing community enablement to connect suppliers digitally to retailers [6][10] Key Points Industry and Market Position - SPS Commerce's total addressable market (TAM) is now estimated at $11 billion, up from a previous estimate of $5 billion, reflecting growth in the EDI market [14][15] - The U.S. accounts for approximately $6.5 billion of the TAM, with a potential global supplier base of about 275,000, including 147,000 in the U.S. [15] Revenue Growth and Customer Acquisition - The company reported 300 net new customers in Q1, indicating a reacceleration in network expansion, driven by community enablement programs [10][12] - Revenue growth can come from new customer acquisition or increasing average revenue per customer, influenced by the nature of community enablement programs [11][12] Pricing Model - The core EDI pricing model is based on the number of connections suppliers have within the network, with a fixed fee per connection making up about 80% of revenue [21][23] - A small variable component is based on the number of documents exchanged, which remains stable despite fluctuations in consumer spending [21][23] Research and Development (R&D) - R&D spending has been consistent at about 10% of sales, focusing on enhancing existing products and internal tools, with a growing emphasis on AI and machine learning [24][26] - The company is also developing rule books to help retailers manage their supply chain processes more effectively [26] New Product Development - SPS recently launched a manufacturing supply chain performance suite aimed at co-packers and manufacturers, addressing upstream supply chain needs [28][29] - The company has also acquired revenue recovery software to help suppliers manage chargebacks and deductions from retailers, presenting a significant cross-selling opportunity [32][36] Financial Metrics and Goals - SPS aims for adjusted EBITDA margins of at least 35%, with current margins in the upper twenties [47] - The company targets gross margins in the low to mid-seventies, with ongoing investments in customer experience expected to drive improvements [48][45] International Growth - Currently, 17% of sales are international, with growth ambitions focused on expanding direct sales and community enablement in Europe following the acquisition of Thai Kinetics [51][52] Data Monetization Opportunities - SPS is exploring ways to monetize the data generated from transactions on its network, potentially offering insights for demand planning and forecasting [73] Additional Insights - The company has a strong focus on community enablement programs, which not only drive supplier connections but also enhance revenue opportunities through existing customer relationships [56] - The integration of recent acquisitions is ongoing, with efforts to standardize pricing and service delivery models [40][41] This summary encapsulates the key points discussed during the SPS Commerce FY Conference Call, highlighting the company's strategic initiatives, market position, and financial outlook.

Revvity (RVTY) FY Conference Summary Company Overview - **Company**: Revvity (formerly PerkinElmer) - **Industry**: Life Sciences Tools and Diagnostics Key Points and Arguments Market Environment and Company Adaptation - The current market is dynamic with uncertainties due to policy changes, tariffs, and challenges in pharma, biotech, and academia sectors [3][4][5] - Revvity's diverse portfolio has demonstrated resilience amidst these challenges, with 60% of revenue now coming from diagnostics and software [4][5][8] Financial Performance and Growth - Revvity's growth rate is at the top end of its publicly traded peer group, with a long-range plan (LRP) growth target of 6-8%, compared to 3-5% previously [4][7] - The company has shifted to a recurring revenue model, with over 80% of revenue now from recurring sources, improving margins from 18-20% to 28% [7][8] - Organic growth guidance for the year is set at 3-5%, with confidence in achieving this despite market uncertainties [9][12] Life Sciences Segment Challenges - The life sciences segment faces challenges due to funding shifts towards clinical work, impacting preclinical discovery [14][15] - Revvity's differentiated product portfolio in consumables is expected to maintain growth despite these challenges [15][20] Software Business - Revvity's software business is unique, functioning as an ERP for researchers, with 48 out of the top 50 pharma companies using its software [26][28] - The software segment is expected to grow significantly, contributing to overall margin improvement [56] Diagnostics and Growth Opportunities - Immunodiagnostics represent a significant growth opportunity in the U.S., with current market penetration at 15-20% compared to a historical 35-40% [30][31] - The company is focusing on expanding its offerings in reproductive health and rare disease testing, leveraging partnerships for growth [40][41] Competitive Landscape in China - Revvity faces intense competition in China, but differentiates itself through proprietary assays and a focus on complex diseases [35][36] - The company is adapting to local market conditions and regulatory environments to maintain growth and profitability [36][39] Capital Allocation and M&A Strategy - Revvity has been active in M&A, completing 13 acquisitions in 22 months to enhance its portfolio [60][62] - Current capital allocation focuses on share buybacks, with a balanced approach to future M&A opportunities [64][68] Long-term Outlook - The company aims for mid-30s operating margins in a normalized market environment, with equal opportunities for margin expansion across life sciences and diagnostics [54][56] - Revvity's strong portfolio and execution strategy position it well for long-term growth despite current macroeconomic challenges [72][74] Additional Important Insights - The company emphasizes innovation and automation in its product offerings to enhance researcher productivity, especially in a budget-constrained environment [47] - Revvity's software and diagnostics segments are seen as critical to its long-term success, with ongoing efforts to improve customer stickiness and market penetration [25][28][70]

什么是算法? 今天，"算法"这个词已为越来越多的人所熟知。从微信上传递文件和图片，到百度、谷歌等搜索引擎上 的网页排名，再到收发电子邮件，我们几乎没有一天不与算法打交道。 脸书很少公开谈论其算法的具体运作方式。事实上，在单个用户的层面上，它自己也不知道是怎么运作 的。算法产生的结果对每个用户来说都是独一无二的，就像他们的指纹一样。 那么，什么是算法呢? 如果你去问计算机科学家，他们会告诉你：算法就是用计算机语言编写的一串指令，它接受输入，对输 入的信息执行一些可重复的运算，然后提供输出。 一个简单的例子是冒泡排序算法(或许你在学计算机语言的时候，自己就编过这样的程序)，你向它输入 一串数字，让它按从小到大的顺序重新排列。它首先比较前两个数字。如果前一个大于后一个，就将它 们对调。否则，就换到下一对。它会一次又一次地循环，直到不需要任何交换为止，这时它就会输出一 个排好序的数字列表。如果你在网上购物时按价格从低到高筛选产品，那么冒泡排序算法就会在幕后启 动。 不过，"算法"一词的流行用法正在发生变化：它越来越多地被用来描述计算机所完成的几乎任何事情。 这也包括人工智能(AI)和机器学习领域，在这些领域，算法的步 ...

Core Points - The third United Nations Ocean Conference, co-hosted by France and Costa Rica, opened in Nice, France on June 9 [1] - The China National Ocean Information Center led a side event titled "Smart Ocean: Innovative Science Leading Action for a Sustainable Future" [1] - The Ministry of Natural Resources of China publicly released the China Global Ocean Fusion Dataset 1.0, which integrates over 40 different data sources and includes China's independent ocean observations [1] Summary by Categories Dataset Features - The China Global Ocean Fusion Dataset (CGOF1.0) has a time span of up to 60 years and a spatial resolution of 10 kilometers [1] - The dataset incorporates advanced AI technologies such as deep learning, transfer learning, and machine learning, resulting in improved accuracy compared to mainstream foreign datasets [1] International Collaboration - The event highlights China's commitment to international collaboration in ocean data sharing and sustainable ocean management [1] - The integration of diverse data sources reflects a global effort to enhance oceanic research and monitoring [1]