CNN

Search documents
建模市场与人机共振:李天成超越价格预测的认知框架
Sou Hu Wang· 2025-06-30 10:40
主讲人:Leonhard 李天成 我们首先要达成一个共识:市场是不可被"精确预测"的。任何声称能预测明日确切涨跌幅的个体或系 统,本质上都游走在随机性与欺诈的边缘。我们的目标,从来不是成为一个水晶球,而是构建一个能够 以更高概率理解市场当前状态(State)并推断其短期演化方向的认知框架(Cognitive Framework)。 交易的本质,是在一个非平稳(non-stationary)、高噪音的随机过程中,寻找期望收益为正的决策机会。 传统的技术分析,无论是形态学还是指标,都是在尝试对这个过程进行降维观测,但这种降维是粗暴且 失真的。它忽略了驱动价格这个一维标量背后,那个高维的、由无数因子构成的潜在空间(Latent Space)。 一、从序列依赖到结构依赖:模型的范式演进 在深度学习的早期应用中,我们尝试用不同的模型去捕捉不同的数据依赖性。 λ_k(t) = μ_k + Σ ∫_0^t g_k,i(t-u) dN_i(u) CNN(卷积神经网络),其核心是卷积核,本质上是一种 局部空间模式(local spatial pattern)的提取器。我 们将时间序列数据转化为图像矩阵(例如,使用Gramian ...
刚刚,何恺明官宣新动向~
自动驾驶之心· 2025-06-26 10:41
来源 | 新智元 点击下方 卡片 ,关注" 自动驾驶之心 "公众号 戳我-> 领取 自动驾驶近15个 方向 学习 路线 >>自动驾驶前沿信息获取 → 自动驾驶之心知识星球 本文只做学术分享,如有侵权,联系删文 【导读】 AI大神何恺明正式入职谷歌DeepMind,担任杰出科学家,同时保留MIT终身副教授身 份。从Meta到MIT,再到如今的谷歌,这位「学界+业界」双修的大牛,将为DeepMind的AGI注 入一针强心剂。 AI圈炸了!CV大牛何恺明正式官宣入职谷歌。 已更新的个人主页上,明确写着:兼职谷歌DeepMind杰出科学家。 与此同时,他依然保留MIT EECS终身教授的身份。 这位CV领域的传奇人物,因提出ResNet而名震江湖,彻底改变了深度学习的发展轨迹,成为现代AI 模型的基石。 如今,这位「学界+业界」双轨并行的跨界大神,再次用行动证明了他的无限可能! 对于谷歌DeepMind而言,何恺明的加入更是如虎添翼。 新人大额优惠!欢迎加入技术最前沿~ ResNet之父再跨界,DeepMind迎超级大脑 何恺明曾是微软亚研院研究院、Meta「明星实验室」FAIR的研究科学家,专注的研究领域包括深度 ...
刚刚,何恺明官宣入职谷歌DeepMind!
猿大侠· 2025-06-26 03:20
编辑:桃子 【导读】 AI大神何恺明正式入职谷歌DeepMind,担任杰出科学家,同时保留MIT终身副教授身 份。从Meta到MIT,再到如今的谷歌,这位「学界+业界」双修的大牛,将为DeepMind的AGI注 入一针强心剂。 AI圈炸了!CV大牛何恺明正式官宣入职谷歌。 已更新的个人主页上,明确写着:兼职谷歌DeepMind杰出科学家。 转自:新智元 与此同时,他依然保留MIT EECS终身教授的身份。 这位CV领域的传奇人物,因提出ResNet而名震江湖,彻底改变了深度学习的发展轨迹,成为现代AI 模型的基石。 如今,这位「学界+业界」双轨并行的跨界大神,再次用行动证明了他的无限可能! 对于谷歌DeepMind而言,何恺明的加入更是如虎添翼。 他的技术专长,涵盖了计算机视觉、深度学习等核心领域,学术影响力在全球范围内有目共睹。 Demis Hassabis曾公开表示,AGI可能在未来5-10年内实现。 何恺明的到来,无疑将助力这一终极目标的加速实现。 ResNet之父再跨界,DeepMind迎超级大脑 何恺明曾是微软亚研院研究院、Meta「明星实验室」FAIR的研究科学家,专注的研究领域包括深度 学习和计 ...
何恺明新身份:谷歌DeepMind杰出科学家
机器之心· 2025-06-26 00:30
机器之心报道 机器之心编辑部 恭喜谷歌。 刚刚,有网友爆料,自己在公司收到了「欢迎何恺明加入」的邮件,何恺明疑似加入谷歌。 在搜索何恺明个人主页后,我们可以确认,他确实加入了谷歌,不过是以兼职的形式,职位是谷歌 DeepMind 杰出科学家 (Distinguished Scientist)。 个人主页:https://people.csail.mit.edu/kaiming/ 关于何恺明在谷歌的具体研究方向,目前还搜不到详细信息。 不过,我们可以根据他最近发表的研究推测一下。前段时间,他所在的团队发布了一篇题为「Mean Flows for One-step Generative Modeling」的论文(参见《 何恺 明团队又发新作: MeanFlow 单步图像生成 SOTA,提升达 50% 》)。在最近的 CVPR workshop 上,他也重点介绍了这一论文所代表的方向。 在分享中,他指出,在 AlexNet 之前,逐层训练更为流行,如深度信念网络(DBN)和去噪自编码器(DAE)。但 AlexNet 之后,识别模型普遍实现了端到端训 练,大大简化了模型设计和训练的复杂性。 不过,有趣的是,今天的生成 ...
微算法科技(NASDAQ:MLGO)采用量子卷积神经网络(QCNN),检测区块链中的DDoS攻击
Zhong Guo Chan Ye Jing Ji Xin Xi Wang· 2025-06-18 02:18
Core Viewpoint - The article discusses the increasing security issues in blockchain technology, particularly focusing on DDoS attacks and how quantum convolutional neural networks (QCNN) developed by Micro Algorithm Technology (NASDAQ: MLGO) can enhance detection and response capabilities against these threats [1][7]. Group 1: Quantum Convolutional Neural Network (QCNN) Development - Micro Algorithm Technology has innovatively improved QCNN for detecting DDoS attacks in blockchain networks by optimizing quantum bit initialization and control methods, enhancing stability and reliability [1][7]. - The structure of QCNN has been adjusted to better handle blockchain transaction data and network status information, making it more suitable for the specific characteristics of blockchain data [1][7]. - Specialized quantum state reading and parsing technologies have been developed to accurately extract features related to DDoS attacks from quantum computation results [1][7]. Group 2: Data Collection and Preprocessing - Data collection involves gathering various types of data from the blockchain network, including transaction data, node status information, and network traffic data, using APIs and monitoring tools [3]. - Preprocessing of collected data is crucial for the effective operation of QCNN, involving data cleaning, noise reduction, and standardization to ensure data quality [3]. - Feature extraction is performed to identify characteristics related to DDoS attacks, such as transaction frequency and network traffic changes, which serve as inputs for the QCNN [3]. Group 3: Quantum Operations - Quantum bit initialization ensures that quantum bits are in a stable initial state, balancing the number of quantum bits with computational complexity [4]. - Quantum convolution operations utilize the properties of quantum bits to extract features and recognize patterns from input data through a series of quantum gate operations [4]. - Quantum pooling operations reduce data dimensions while retaining important features, employing a measurement-based pooling method to select the most probable quantum states [5]. Group 4: Classification and Output - After quantum convolution and pooling, a quantum fully connected layer processes the low-dimensional quantum state for DDoS attack classification and detection [6]. - The output from the quantum fully connected layer is a quantum state representing classification results, which is converted into a readable format using specialized quantum state reading techniques [6]. - If the probability distribution indicates a high likelihood of a DDoS attack, alerts are generated to notify network administrators for appropriate defensive actions [6]. Group 5: Applications and Future Prospects - The QCNN developed by Micro Algorithm Technology can monitor blockchain networks in real-time, promptly detecting signs of DDoS attacks and issuing alerts for immediate defensive measures [7]. - This technology can be integrated with other security measures, such as encryption and access control, to create a more secure blockchain environment [7]. - As quantum computing technology advances, the application prospects for QCNN in detecting DDoS attacks will expand, potentially enhancing computational power and accuracy [7].
Warner Bros. Discovery Is Splitting Up: What It Means for You
CNET· 2025-06-09 16:37
Core Points - Warner Bros. Discovery is splitting into two separate public companies: Streaming & Studios and Global Networks [2][4] - The split is expected to be completed by 2026, following the merger that occurred in 2022 [4] - Streaming & Studios will include HBO Max, Warner Bros. movies, gaming, and DC, while Global Networks will encompass Discovery Plus, CNN, Bleacher Report, and TNT Sports [3] Company Structure - Streaming & Studios will focus on streaming services and studio operations, including the newly renamed HBO Max [3] - Global Networks will manage traditional media assets and networks, including CNN and Discovery Plus [3] Consumer Impact - It remains unclear how the split will affect existing subscribers, particularly regarding content access and potential pricing changes [4][5] - Current services are not expected to undergo major changes immediately, with a focus on shareholder value and new ventures [5]
Warner Bros. Discover Is Splitting Up: What It Means for You
CNET· 2025-06-09 15:59
Core Points - Warner Bros. Discovery is splitting into two separate public companies: Streaming & Studios and Global Networks [2][4] - Streaming & Studios will encompass HBO Max, Warner Bros. movies, gaming, and DC properties, while Global Networks will include Discovery Plus, CNN, Bleacher Report, and TNT Sports [3] - The split is expected to be completed by 2026, following the merger that occurred in 2022 [4] Company Impact - The split may create confusion among streaming customers due to the generic nature of the new company names [2] - There is uncertainty regarding whether the split will affect consumer access to content on existing subscriptions, such as HBO Max [4] - Current services are not anticipated to undergo major changes, with a focus on shareholder value and new ventures rather than customer impact [5]
Warner Bros. Discovery announces major corporate restructuring to separate streaming from cable
Fox Business· 2025-06-09 15:36
Group 1 - Warner Bros. Discovery (WBD) will split into two companies, separating its studios and streaming business from its cable TV networks to enhance competitiveness in the streaming market [1][5] - CEO David Zaslav will lead the streaming and studios business post-split, while CFO Gunnar Wiedenfels will oversee the global networks unit, aiming for sharper focus and strategic flexibility [2] - The split is structured as a tax-free transaction expected to be completed by mid-2026, with WBD shares rising by 8% during morning trading [5] Group 2 - The corporate split follows the 2022 merger of WarnerMedia and Discovery and aligns WBD with Comcast's strategy of spinning off cable TV networks [5][6] - WBD has initiated tender offers to restructure its existing debt, supported by a $17.5 billion bridge facility from JPMorgan, with plans to refinance before the separation [9] - The global networks division will retain up to a 20% stake in the streaming and studios business, which it intends to monetize to further reduce debt [9]
Warner Bros Stock Surges on Company Split
Schaeffers Investment Research· 2025-06-09 15:05
Media & entertainment stock Warner Bros Discovery Inc (NASDAQ:WBD) was last seen up 10.1% at $10.81, after the company announced it will split into two publicly traded companies by next year. CEO David Zaslav will head the streaming and studios company, which will contain HBO Max, while CFO Gunnar Wiedenfels will become CEO of the global networks business, which includes CNN, TNT Sports, and Discovery.Today's pop has Warner Bros stock breaking above recent pressure at $10, trading at its highest level since ...
华纳兄弟探索(WBD.US)拟分拆流媒体与有线电视 迎战奈飞
智通财经网· 2025-06-09 13:19
智通财经APP获悉,华纳兄弟探索公司(WBD.US)宣布,将在明年年中将公司分拆为两家独立上市公 司,实现流媒体及影视制作业务与电视网络业务的剥离。 华纳兄弟同时宣布,将筹集175亿美元的过渡贷款,预计在分拆前完成资本重组。全球网络公司还将持 有流媒体和影视制作业务至多20%的股份,并通过套现减持以削减债务。 作为CNN、HBO等频道的母公司,华纳兄弟于2022年由AT&T旗下华纳媒体(WarnerMedia)与探索公司 (Discovery Inc.)合并而成。这笔交易催生的新公司背负沉重债务,而其最大业务有线电视正面临观众和 广告收入的双重流失。 上周五,该公司美股收涨1.8%至9.82美元,今年以来累计下跌7.1%。截至发稿,该股周一盘前上涨 8%。 该公司周一在声明中表示,流媒体和影视制作公司将涵盖华纳兄弟电视、华纳兄弟电影集团、DC工作 室、HBO及HBO Max平台,以及影视库内容,由首席执行官大卫·扎斯拉夫(David Zaslav)领导。首席财 务官冈纳·维登费尔斯(Gunnar Wiedenfels)将执掌新成立的全球网络公司,业务包括CNN等娱乐、体育和 新闻电视品牌。 华纳兄弟近期刚完成 ...