CP3405 CP2408 Lean UX

ORB-SLAM3 bool Tracking::NeedNewKeyFrame() 是 ORB-SLAM3 的关键帧插入决策器IMU 未初始化:每 0.25s 强制插入,快速积累初始化数据;仅跟踪/建图停止/重定位冷却:直接拒绝;常规判定:时间、空闲、跟踪弱;c2:当前内点相对参考关键帧不足,或近点覆盖不足;二者与运算决定视觉上是否需要新关键帧;c3:IMU 系统 0.5s 强制插入;c4:单目惯性低匹配/丢失恢复;LocalMapper 仲裁:空闲直接插入;忙则中断 BA,双目/RGBD 队列 <3 可插入,单目更严格。它体现了 ORB-SLAM3 在地图密度、跟踪鲁棒性、计算负载。 阅读详情


CP3405 Starting-point Document
Excerpts from a CP2408 Lean UX Project
This document is the foundation for your CP3405 project this trimester. The following sections describe the Lean UX iterations completed by a previous CP2408 team, which you were not part of before. It aims to give you the experience of continuing an ongoing project rather than starting anew, likely a new challenge!
Please refer to Assessment Task 3 Part 1 for details about the first part of your CP3405 Scrum project, the Pilot Study.
Iteration 1
Title of Initiative: Augmented Reality Educational Game Development Date: Oct 6th
1. Business Problem
• Problem Statement: Traditional educational methods often fail to maintain
student engagement, especially given digital distractions.
• Market Trends: Growing integration of interactive technology like AR in education.
• Customer Behaviour: Increasing demand for engaging and interactive educational
tools.
2. Business Outcomes
• Success Metrics: Number of teachers interested in participating in further
development, quality of feedback collected, initial indicators of teacher and
potential student engagement.
• Indicators of Success: Positive feedback on the concept of AR in classrooms, the
willingness of teachers to explore AR tools, and actionable suggestions for AR
content.
3. Users
• Primary Users: Teachers at middle and high school levels.
• Secondary Users: Educational administrators and students.
4. User Outcomes & Benefits (Anticipated)
• Teacher Benefits: Access to innovative tools that increase student engagement,
enhanced ability to convey complex concepts, and professional development in
modern educational technologies.
• Student Benefits: Improved engagement and comprehension through interactive
AR learning experiences.
5. Solutions
• Initial Approach: Create a detailed presentation to educate and engage teachers
about the benefits of AR in education.
• Engagement Strategy: Conduct interviews with teachers post-presentation to
gather their insights and requirements.
6. Hypotheses

• Hypothesis Statement: "We believe that teacher adoption of AR tools in the classroom will increase if they understand the benefits and participate in the development process."
7. What’s the most important thing we need to learn first?
• Key Assumption: Teachers are open to using AR technology and see its potential
value in enhancing student engagement.
• Focus Area: Teacher familiarity with AR, openness to technological integration,
and specific educational needs that AR could address.
8. What’s the least amount of work we need to do to learn the next most important thing?
• Minimal Action: Develop and present an informative session about AR applications in education to a group of teachers.
• Evaluation Approach: Use surveys and interviews post-presentation to collect qualitative and quantitative data on teacher responses.
Outcomes of Iteration 1 MVP (presentation and interviews)
 Conceptual Approval:
• Teachers were excited about AR’s potential to make learning about the solar
system more engaging, especially through the demonstration of planetary
distances.
• Teachers appreciated the idea of using AR to help students visualise the vast
distances between the planets and the sun, noting how this could address common
misconceptions about the solar system's scale.  Desired Features:
• Focus on Scale and Distance: Teachers emphasised the importance of accurately depicting the distances between planets. They felt that an AR tool that could effectively scale these distances would be far more engaging than static images or traditional classroom models.
• Interactivity: Teachers wanted students to be able to interact with the AR models by zooming in and out to explore the vast distances and get a sense of scale.
 Educational Content:
• Layered Information on Distances: Teachers suggested starting with an overview
of the entire solar system and allowing students to zoom in on individual planets while showing the relative distances. They stressed that the distances should be visually impactful and scientifically accurate.
 Classroom Integration:
• Lesson Plans: Teachers expressed the need for lesson plans specifically focused on
using AR to demonstrate planetary distances. These plans should guide how to incorporate the AR tool into classroom activities, making it easier to teach the scale of the solar system.
• Guided Exploration: Teachers requested that the AR tool include a guided exploration feature that walks students through the distances between the planets, highlighting interesting facts about their orbits and positioning in the solar system.
 Technical Considerations:

• Ease of Use: Teachers reiterated that the AR tool should be easy to set up and use in the classroom. Minimal setup time was a key requirement to ensure smooth integration into lessons.
• Compatibility: The tool should work across commonly available devices in schools, ensuring that no students are left out due to hardware limitations.
 Feedback and Iteration:
• Pilot Program: Teachers recommended starting with a small pilot program that
focuses on solar system distances to test how well the AR tool enhances student
understanding of this specific concept.
• Continued Feedback: Teachers encouraged ongoing feedback collection during the
pilot to ensure the AR tool evolves based on real classroom experiences and insights into how well it communicates planetary distances.
Iteration 2
Title of Initiative: Augmented Reality Educational Game Development Date: Oct 20th
1. Business Problem:
o Problem Statement: Teachers need an effective and engaging tool to demonstrate the relative distances between the planets and the sun, which traditional teaching methods often fail to illustrate in a meaningful and realistic way.
o Market Trends: Increasing demand for immersive and interactive educational tools, especially in astronomy and science education.
o Customer Behavior: Teachers are seeking innovative ways to improve student comprehension of abstract concepts, like astronomical distances, through AR technology.
2. Business Outcomes:
o Success Metrics: Quality of feedback on how effectively the AR tool represents
planetary distances, clarity in understanding teachers’ educational needs, and
readiness to move toward a more complex prototype.
o Indicators of Success: Teachers can clearly visualise the potential of AR to
represent planetary distances, express confidence in using such a tool for education, and provide actionable feedback on AR content for future development.
3. Users:
o Primary Users: Teachers at middle and high school levels teaching astronomy. o Secondary Users: Educational administrators evaluating the tool for broader
curriculum integration.
4. User Outcomes & Benefits (Anticipated):
o Teacher Benefits:
 A tool to help explain the vast distances between planets in a more
intuitive and interactive way.
 Improved confidence in teaching difficult concepts through innovative
technology.

 Opportunity for professional development in modern educational technologies.
5. Solutions:
o Current Approach: Create paper prototypes and cardboard models to illustrate
how AR will represent the relative distances between planets and the sun.
These models will be used in interactive sessions with teachers.
o User Interaction Focus: Use the paper prototype to gather teacher feedback
on the effectiveness of the AR simulation, specifically focusing on the accurate
representation of distances. 6. Hypotheses:
o Hypothesis Statement: "We believe that teachers will find AR simulations of planetary distances more effective and engaging for teaching astronomy concepts if they experience a realistic representation of these distances."
7. What’s the most important thing we need to learn first?:
o Key Assumption: Teachers will find value in using AR to depict the scale of the
solar system and believe it improves students’ conceptual understanding.
o Focus Area: Validate whether the AR representation of distances aligns with
teachers' needs and expectations.
8. What’s the least amount of work we need to do to learn the next most important
thing?:
o Minimal Actionable Step: Develop a paper prototype to visualise the distances
between planets, focusing on scale and relativity. Test with teachers to
evaluate engagement and feedback.
o Evaluation Method: Structured interviews, surveys, and observations during
the interactive prototype testing sessions to capture feedback on AR’s potential impact on teaching planetary distances.
Outcomes of Iteration 2 MVP (paper prototype)
 Engagement with the Concept of Distance:
• "The prototype does a great job of visually representing the relative distances
between planets. I immediately saw how this could capture students' attention in a
way that traditional diagrams can't."
• "The physical interaction of moving planets around gave me a better understanding
of how the AR tool would work in a classroom. It’s an excellent way to simplify a
very abstract concept like astronomical distances."  Visual Representation of Scale:
• "While the paper model is effective in showing relative distances, we need to ensure that in the AR version, the distances between planets feel even more dramatic. Students often struggle to grasp just how far apart these celestial bodies are."
• "The concept of scale was well-represented, but there’s a need for clear markers or reference points, perhaps using familiar objects or distances, so students can truly grasp the magnitude of the solar system."
 Educational Impact:

• "I can see how this tool could be a game-changer for our class work. Students often have trouble visualising how the planets are spaced, and this could bridge that gap, making the content more accessible and engaging."
• "This would certainly help address the misconception many students have about the uniform spacing of planets. The AR tool could help them understand that distances between planets vary dramatically."
 Potential Improvements:
• "It would be great if the AR simulation could incorporate a feature that shows
planets in motion, so students can see how the distances change as they orbit the
sun. That could make the experience even more dynamic."
• "We need to make sure that the AR tool can show both a zoomed-in view for detail
and a zoomed-out view for scale. Switching between these perspectives could help students understand both the planets individually and their relationships to one another."
 Ease of Use and Classroom Integration:
• "The prototype seems simple enough to use, and I imagine the AR tool would be
intuitive as well. But we’ll need to ensure that it integrates easily into existing
lesson plans without requiring too much setup or technical knowledge."
• "One suggestion is to include a guided mode where the AR tool leads students
through the distances between planets step by step. This would be particularly
useful for self-paced learning or group work."  Additional Features:
• "I would love to see more interactive features, like quizzes built into the AR tool, where students could test their knowledge of planetary distances after exploring the model."
• "A feature that allows students to compare planetary distances to everyday distances on Earth, like the distance between cities, could help make the concept even more relatable."
 Overall Feedback:
• "This is a great starting point. The prototype effectively demonstrates the
potential of AR to enhance teaching about the solar system. I’m excited to see how
this could evolve into a digital version that students can engage with in real-time."
• "The idea of using AR to teach planetary distances is very promising. I think this
tool will help make complex astronomical concepts more approachable and engaging for students, and I’m eager to see the next iteration."

 

船载 OpenHarmony AIS 航海工作台:网络层与后端 API 通信 本文介绍船载OpenHarmony AIS航海工作台的网络层设计,通过单例HttpManager统一管理HTTPS请求,采用应用标识、随机串与密钥的MD5签名实现安全鉴权,密钥构建注入不落地。接口涵盖天气、潮汐、水位等,响应统一结构并确保资源释放。关键数据如潮汐采用“缓存优先”策略,支持离线查看。配套管理器协同处理多类网络任务,保障弱网体验与系统稳定。 阅读详情

相关推荐

图谱后端选型:Memgraph vs Neo4j vs 嵌入式 SQLite,什么时候该切换?

本文通过脚本自动探测代码图谱规模,对比SQLite、Memgraph与Neo4j在不同节点量级下的性能表现。实测表明:10万节点内嵌入式SQLite凭借低延迟和轻量部署仍具竞争力;超15万节点后,Memgraph适合低延迟、高并发场景;而Neo4j仅在百万级图谱且需复杂查询时才具备优势。配套提供一键测试延迟与平滑迁移后端的Python工具链,实现按需选型与无缝切换。

xianyinsuifeng的博客 61

可靠性设计全解析

来源定义要点ISO/IEC 25010(软件质量模型)系统在指定条件下、指定时间段内执行规定功能的能力,包含容错性、可用性、可恢复性、成熟性等子特性Google SRE系统在给定时间内正常工作的概率,通常用可用性(Availability)与 SLO 表达工作负载在预期时间内正确、一致地执行其预期功能的能力,包括从基础设施或服务中断中恢复的能力传统可靠性工程(RAMS)可靠性 R、可用性 A、可维护性 M、安全性 S 的综合,强调全生命周期。

流水账 614

船载 OpenHarmony AIS 航海工作台:项目概览与系统定位

本文介绍一款基于OpenHarmony的船载AIS航海工作台,定位为RK3588终端上的常驻桌面应用(Launcher),集成本船定位、AIS监控、海图导航、防碰撞告警、气象潮汐、通信多媒体及系统运维等能力。采用Kiosk模式全屏运行,保障实时性与离线可用性,通过串口接入GNSS/AIS数据,对接后端服务实现OTA升级与接口鉴权,核心设计遵循实时优先、离线可用、安全脱敏三大原则,构建一体化船载智能航海平台。

专注于计算机研发知识和资讯分享 33

从邓明学习的后端笔记

在这里面对于高性能和高可用的支持是更多的,因为高并发可能相对更困难,成本更高,高性能的支持比如不要使用跨页查询,由业务维护之前搜索的位置,使用索引的方式比如 MongoDB 查询的没有索引使用索引,高可用比如分区存储多份且同一个的不同副本在不同的机器。可靠性包括三部分第一个生产者发送的没有正常被接收,可以通过 ack 和回表查询解决,第二个是 bRokeR 宕机还没持久化消息丢失,解决方法可以设置持久化频率,第三个是消费者异步处理的时候没有处理完成就标记已经消费,解决方法可以是处理完再移动偏移量。

m0_57222081的博客 245

从零理解Brep:用Python玩转3D建模的边界表示法(附OpenCASCADE示例)

本文深入解析了边界表示法(Brep)在3D建模中的应用,通过Python和OpenCASCADE的实战示例,详细介绍了Brep的核心数据结构、布尔运算、网格转换及3D打印优化技巧。适合CAD设计、3D打印和计算机图形学领域的开发者学习与实践。

n4o5p6q7r的博客 434

qmc5883p芯片坐标系方向定义

问题 最近一个项目用到qmc5883p, 然而确定磁力计方向的时候发现datasheet上写的不是很清楚, 并且example中还有一个设置方向的一步, 于是测试了芯片坐标方向并记录了下来; 该芯片和qmc5883L方向并不一致; 结果 qmc5883P的datasheet给出来如下说明, Y为笔者添加的; 但是这个方向是按照datasheet的example设置的结果: //Define the sign for X Y and Z axis qmc5883p_write_reg(0x2

时有限 4122

SpringBoot 新闻推荐系统:88个Java文件105个接口,首页推荐位是后端按点击数排出来的

# SpringBoot 新闻推荐系统:88个Java文件105个接口,首页推荐位是后端按点击数排出来的 > 这是一套基于 Spring Boot 的新闻推荐系统源码,配套 `db.sql` 建表脚本,业务覆盖新闻发布、排行榜管理、评论回复、点赞点踩、收藏夹和轮播图配置。形态是**前后台双端**:前台是读者用的 layui 静态站,后台是管理员用的 Vue 单页应用(编译产物已嵌在工程里)。最值

面朝大海 882

Qwen3.6-27B本地部署实战:LM Studio配置、硬件门槛与避坑指南

大语言模型本地部署是当前AI工程落地的核心能力之一,其本质是将高参数量模型(如27B级)在受限硬件环境下实现低延迟、高稳定推理。关键技术原理涵盖量化压缩(如GGUF格式与q4_k_m精度权衡)、GPU显存分配策略(KV Cache管理与Offloading机制)、以及运行时环境兼容性(CUDA驱动、llama.cpp后端版本)。该能力直接决定私有化AI应用的技术可行性与成本边界,广泛应用于企业知识库构建、代码辅助、长文档分析等对数据主权和响应确定性要求高的场景。本文聚焦Qwen3.6-27B这一典型27B模

weixin_33272631的博客 469

BLG大战T1导致应用大规模启动

2026年11月9日,BLG对战T1赛事结束后4小时,因iOS灵动岛功能特性引发大规模应用后台启动。赛事期间用户开启灵动岛,赛后系统推送结束消息,但锁屏UI延迟4小时集中移除,触发大量客户端后台唤醒,导致登录态续期、设备指纹等服务告警。经排查,问题仅限iOS端,根源为灵动岛默认4小时后移除锁屏界面时自动唤醒应用。优化措施包括:临时扩容、推送打散(多通道串行+时间错峰)、客户端识别并保护后端请求,有效缓解压力。

weixin_42448366的博客 209

R语言clusterProfiler实战:GO与KEGG富集分析从入门到精通

本文提供了一份详尽的R语言clusterProfiler实战指南,从环境配置、数据导入、基因ID转换到GO与KEGG富集分析及可视化,手把手教你解读生物学意义。文章重点介绍了如何利用clusterProfiler这一强大工具,将差异基因列表转化为清晰的生物学通路与功能故事,并分享了结果解读、可视化以及处理非模式生物等进阶技巧与避坑指南。

weixin_29044713的博客 162

ComfyUI的G-Dino模型加载器报错解决方案:从崩溃到重生

遇到G-Dino模型加载器报错不要慌,按照本文的步骤操作,99%的问题都能解决。模型文件正确下载到指定位置segment_anything节点安装完整且无冲突需要补充更多细节或故障排除方法吗?可以随时告诉我。

2301_78150483的博客 4024

带着小白学习-微服务-6 实战:网关

地址栏输入:http://localhost:8088/goods/index,这时候就不是到8100端口去调用了,而是8088直接通过路由转发,分别到nacos对应的goods对应的服务中区调用。** 可以看出,这里将前面已经正常运行的server-store子工程(库存服务,服务提供方),server-goods子工程(商品服务,服务消费方)。(1)uri: lb://store 代表要到nacos的注册中心的服务列表中找到服务名为store的微服务;,网关是做路由转发,不是服务调用;

wdyan297的博客 250

激光雷达的64线,32线,16线,你弄懂了吗?

激光雷达的 64 线、32 线、16 线是指激光雷达系统中能够同时发射和接收激光脉冲的激光线的数量5。

谢谢大家的关注和点赞!这里只有纯纯的知识干货,没有一句废话。希望能实实在在帮到大家~要是觉得有用,别忘了给我点支持哟,你 2641

HTML+CSS+JS对话式实战(5)-活动管理页之查看报名名单的实现

本文实现活动管理中的“查看报名名单”功能,通过POST接口 /activity/getSignList 根据活动ID查询报名信息,返回统一格式数据。前端在活动列表页新增【查看报名】按钮,点击后弹出独立窗口展示报名人员列表,支持空数据提示。使用 postRequest 封装请求,动态渲染表格,逻辑清晰可扩展。测试验证了正常、空名单及错误场景,最终优化弹窗样式,提升用户体验。

编程副队长 219

SAP ABAP Append structure 介绍

所有在 append structure 上增强的字段都是 append structure 的一部分,这意味着对这些增强所做的任何更改仅仅施加在 append structure 上,而不会影响被增强的 ABAP DDIC 结构。这同样适用于 transport. 激活数据库表或结构时,会找到所有关联的 append structure,并将它们的组件附加到数据库表或结构。Append structure(附加结构)是 ABAP 字典中附加到另一个结构或数据库表并将其组件附加到它们的结构。

2007 年 ~ 2025 年,深耕 SAP 技术 18 年 1168

Spring AI Alibaba实战训练营-16 NL2SQL (自然语言转 SQL)应用开发指南

本文介绍了一个基于AI的智能BI应用开发项目,主要实现自然语言转SQL(NL2SQL)功能,包含对话模块、MCP模块、向量数据管理模块。项目采用Spring AI Alibaba框架,集成通义千问大模型,支持内存向量库和AnalyticDB向量数据库两种存储方案。重点展示了chat模块的三个控制器实现:AnalyticNl2SqlController用于生产环境,SimpleChatController适用于开发测试,Nl2sqlForGraphController支持基于图的复杂对话流程。系统通过将业务知

program哲学 1590

【AI】Agent 全栈进阶|AgentScope 沙箱源码解析

本文基于AgentScope源码,解析沙箱机制在DockerWorkspace中的实现原理。沙箱通过隔离执行环境,限制Agent对宿主机的副作用,保障安全;统一根路径/workspace,支持技能、会话与数据隔离,并借助MCP Gateway实现容器内进程通信。核心设计包括Backend抽象层、生命周期模板、目录结构规范及中转通信机制,实现权限与隔离解耦,适用于生产级Agent安全运行场景。

qq_45797138的博客 357

VC中sqlite的使用

在VC中对sqlite的使用,实现数据库大批数据的快速插入,读取数据库中的数据在ListCtrl中显示

MCP实战手记(十四):MCP调用链追踪,`_meta`注入traceparent的独家方案

篇13已在网关补齐多租户,本篇解决MCP Client→Server链路不可见问题:SkyWalking、OTel Java Agent识别不了`_meta`通道。手动向`_meta`注入 W3C traceparent传递链路,给出可运行代码,支持注入、续接 span、下游透传;Micrometer 埋点延迟、QPS、成功率、P95、错误分布,日志携带traceId与租户。附选型与Docker 栈,实测MCP Java SDK 2.0.0透传可用,支持生产落地。

weixin_39885962的博客 1094

C# WinForm 基于DirectShow.dll编写的程序,可边预览边保存视频,音频和图片

基于微软的DirectShow.dll的二次开发1.自动识别连接的摄像头跟麦克风且可以对其进行选择(默认选中第一个)2.可录制wmv和avi两种格式的视频并且都带有音频,avi可以录制的分辨率和帧数进行选择,wmv默认进行了淸晰度优化如不喜欢可去除代码中有相关提示3.可在预览点开之后进行拍照,也可边录制视频边拍照4.添加了视频和音频的压缩功能(酌情使用)这次二次开发发现C#对DirectShow的相关信息太少,这次做完也不敢藏私共享出来,还望大家指正

efm32 入门.pdf

本PDF共计22页,介绍的是操作驱动API,下载与否自己斟酌。

Vue Router

介绍vue-router的基本使用以及核心概念

k_5412528的博客 224

深度学习工具包 DeepLearningToolbox

深度学习工具包Deprecation notice.-----This toolbox is outdated and no longer maintained.There are much better tools available for deep learning than this toolbox, e.g. [Theano](http://deeplearning.net/software/theano/), [torch](http://torch.ch/) or [tensorflow](http://www.tensorflow.org/)I would suggest you use one of the tools mentioned above rather than use this toolbox.Best, Rasmus.DeepLearnToolbox================A Matlab toolbox for Deep Learning.Deep Learning is a new subfield of machine learning that focuses on learning deep hierarchical models of data.It is inspired by the human brain's apparent deep (layered, hierarchical) architecture.A good overview of the theory of Deep Learning theory is[Learning Deep Architectures for AI](http://www.iro.umontreal.ca/~bengioy/papers/ftml_book.pdf)For a more informal introduction, see the following videos by Geoffrey Hinton and Andrew Ng.* [The Next Generation of Neural Networks](http://www.youtube.com/watch?v=AyzOUbkUf3M) (Hinton, 2007)* [Recent Developments in Deep Learning](http://www.youtube.com/watch?v=VdIURAu1-aU) (Hinton, 2010)* [Unsupervised Feature Learning and Deep Learning](http://www.youtube.com/watch?v=ZmNOAtZIgIk) (Ng, 2011)If you use this toolbox in your research please cite [Prediction as a candidate for learning deep hierarchical models of data](http://www2.imm.dtu.dk/pubdb/views/publication_details.php?id=6284)```@MASTERSTHESIS\{IMM2012-06284, author = "R. B. Palm", title = "Prediction as a candidate for learning deep hierarchical models of data", year = "2012",}```Contact: rasmusbergpalm at gmail dot comDirectories included in the toolbox-----------------------------------`NN/` - A library for Feedforward Backpropagation Neural Networks`CNN/` - A library for Convolutional Neural Networks`DBN/` - A library for Deep Belief Networks`SAE/` - A library for Stacked Auto-Encoders`CAE/` - A library for Convolutional Auto-Encoders`util/` - Utility functions used by the libraries`data/` - Data used by the examples`tests/` - unit tests to verify toolbox is workingFor references on each library check REFS.mdSetup-----1. Download.2. addpath(genpath('DeepLearnToolbox'));Example: Deep Belief Network---------------------```matlabfunction test_example_DBNload mnist_uint8;train_x = double(train_x) / 255;test_x = double(test_x) / 255;train_y = double(train_y);test_y = double(test_y);%% ex1 train a 100 hidden unit RBM and visualize its weightsrand('state',0)dbn.sizes = [100];opts.numepochs = 1;opts.batchsize = 100;opts.momentum = 0;opts.alpha = 1;dbn = dbnsetup(dbn, train_x, opts);dbn = dbntrain(dbn, train_x, opts);figure; visualize(dbn.rbm{1}.W'); % Visualize the RBM weights%% ex2 train a 100-100 hidden unit DBN and use its weights to initialize a NNrand('state',0)%train dbndbn.sizes = [100 100];opts.numepochs = 1;opts.batchsize = 100;opts.momentum = 0;opts.alpha = 1;dbn = dbnsetup(dbn, train_x, opts);dbn = dbntrain(dbn, train_x, opts);%unfold dbn to nnnn = dbnunfoldtonn(dbn, 10);nn.activation_function = 'sigm';%train nnopts.numepochs = 1;opts.batchsize = 100;nn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.10, 'Too big error');```Example: Stacked Auto-Encoders---------------------```matlabfunction test_example_SAEload mnist_uint8;train_x = double(train_x)/255;test_x = double(test_x)/255;train_y = double(train_y);test_y = double(test_y);%% ex1 train a 100 hidden unit SDAE and use it to initialize a FFNN% Setup and train a stacked denoising autoencoder (SDAE)rand('state',0)sae = saesetup([784 100]);sae.ae{1}.activation_function = 'sigm';sae.ae{1}.learningRate = 1;sae.ae{1}.inputZeroMaskedFraction = 0.5;opts.numepochs = 1;opts.batchsize = 100;sae = saetrain(sae, train_x, opts);visualize(sae.ae{1}.W{1}(:,2:end)')% Use the SDAE to initialize a FFNNnn = nnsetup([784 100 10]);nn.activation_function = 'sigm';nn.learningRate = 1;nn.W{1} = sae.ae{1}.W{1};% Train the FFNNopts.numepochs = 1;opts.batchsize = 100;nn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.16, 'Too big error');```Example: Convolutional Neural Nets---------------------```matlabfunction test_example_CNNload mnist_uint8;train_x = double(reshape(train_x',28,28,60000))/255;test_x = double(reshape(test_x',28,28,10000))/255;train_y = double(train_y');test_y = double(test_y');%% ex1 Train a 6c-2s-12c-2s Convolutional neural network %will run 1 epoch in about 200 second and get around 11% error. %With 100 epochs you'll get around 1.2% errorrand('state',0)cnn.layers = { struct('type', 'i') %input layer struct('type', 'c', 'outputmaps', 6, 'kernelsize', 5) %convolution layer struct('type', 's', 'scale', 2) %sub sampling layer struct('type', 'c', 'outputmaps', 12, 'kernelsize', 5) %convolution layer struct('type', 's', 'scale', 2) %subsampling layer};cnn = cnnsetup(cnn, train_x, train_y);opts.alpha = 1;opts.batchsize = 50;opts.numepochs = 1;cnn = cnntrain(cnn, train_x, train_y, opts);[er, bad] = cnntest(cnn, test_x, test_y);%plot mean squared errorfigure; plot(cnn.rL);assert(er<0.12, 'Too big error');```Example: Neural Networks---------------------```matlabfunction test_example_NNload mnist_uint8;train_x = double(train_x) / 255;test_x = double(test_x) / 255;train_y = double(train_y);test_y = double(test_y);% normalize[train_x, mu, sigma] = zscore(train_x);test_x = normalize(test_x, mu, sigma);%% ex1 vanilla neural netrand('state',0)nn = nnsetup([784 100 10]);opts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samples[nn, L] = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.08, 'Too big error');%% ex2 neural net with L2 weight decayrand('state',0)nn = nnsetup([784 100 10]);nn.weightPenaltyL2 = 1e-4; % L2 weight decayopts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samplesnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex3 neural net with dropoutrand('state',0)nn = nnsetup([784 100 10]);nn.dropoutFraction = 0.5; % Dropout fraction opts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samplesnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex4 neural net with sigmoid activation functionrand('state',0)nn = nnsetup([784 100 10]);nn.activation_function = 'sigm'; % Sigmoid activation functionnn.learningRate = 1; % Sigm require a lower learning rateopts.numepochs = 1; % Number of full sweeps through dataopts.batchsize = 100; % Take a mean gradient step over this many samplesnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex5 plotting functionalityrand('state',0)nn = nnsetup([784 20 10]);opts.numepochs = 5; % Number of full sweeps through datann.output = 'softmax'; % use softmax outputopts.batchsize = 1000; % Take a mean gradient step over this many samplesopts.plot = 1; % enable plottingnn = nntrain(nn, train_x, train_y, opts);[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');%% ex6 neural net with sigmoid activation and plotting of validation and training error% split training data into training and validation datavx = train_x(1:10000,:);tx = train_x(10001:end,:);vy = train_y(1:10000,:);ty = train_y(10001:end,:);rand('state',0)nn = nnsetup([784 20 10]); nn.output = 'softmax'; % use softmax outputopts.numepochs = 5; % Number of full sweeps through dataopts.batchsize = 1000; % Take a mean gradient step over this many samplesopts.plot = 1; % enable plottingnn = nntrain(nn, tx, ty, opts, vx, vy); % nntrain takes validation set as last two arguments (optionally)[er, bad] = nntest(nn, test_x, test_y);assert(er < 0.1, 'Too big error');```[![Bitdeli Badge](https://d2weczhvl823v0.cloudfront.net/rasmusbergpalm/deeplearntoolbox/trend.png)](https://bitdeli.com/free "Bitdeli Badge")

QDQ 量化:从「一格代表多大」到图上的显式精度边界

QDQ并非新量化算法,而是将量化参数(步长、零点)以节点形式显式嵌入ONNX图中,实现量化边界可视化与模型可移植。其核心仍为整数 = round(小数 ÷ 步长) + 零点,通过QuantizeLinear与DequantizeLinear节点标记量化位置。虽图上运行仍为FP32,但后端可识别并融合为真INT8计算(如QConv),实现存储压缩与加速。相比隐式缓存,QDQ提升调试性与跨平台兼容性,是量化落地的“白盒”交付方式。

飞鹰技术 178

零后端启动的Vue3中后台模板

Antdv Next Vue Admin 是一套基于 Vue 3.5 + TypeScript 的企业级中后台模板,主打“零后端启动”与“开箱即用”。通过进程内 Nitro Mock 实现无需后端即可运行,内置请求层、CRUD 状态机、权限指令、分片上传等高频能力,支持子包独立复用。其轻量可拆解的架构设计,结合 Tailwind CSS v4 与 reka-ui/shadcn-vue,显著降低学习成本与开发负担,特别适合中大型项目快速落地。

overFlowzy的博客 310

Java 项目实战:交友项目后台系统——网关统一鉴权与 B 端管理后台

本次改造通过引入SpringCloud Gateway实现网关统一鉴权与路由,新建mf-admin管理端复用C端域服务(Dubbo调用),实现管理员登录与用户管理四件查询。核心套路:网关白名单+双体系JWT(同工具不同payload)、管理端零直连业务库、分页形态区分(MySQL自动count/Mongo手动count)、浏览器端独立聚合。踩坑12条,附自检清单与后续改造方向,可复用于多角色后台系统。

Wang的专栏 63

【全栈面试】Excel 导入流程

本文系统梳理全栈工程师面试中关于Excel导入功能的五大核心考察维度:架构分层(DTO/BO/DO职责分离)、前后端校验分工、性能优化(异步+流式读取+批量插入)、异常处理策略(精准反馈与错误报告)、技术选型原理(EasyExcel vs POI)。强调全栈思维应贯穿“契约先行、解耦设计、安全兜底、体验优化”全流程,通过实例展示如何从同步阻塞到异步高效重构,实现高可用、可维护、用户体验佳的完整链路设计。

qq_38980678的博客 311

JSON 驱动完全控制布局

JSON 驱动完全控制布局

ZQ200720的博客 193

山东微程科技|轻量化实体门店数字化系统设计思路

考虑到线下商户网络不稳定场景,增加本地缓存 + 异步队列机制,弱网环境下支持订单暂存,网络恢复后自动同步,减少丢单问题。系统整体采用前后端分离架构,底层基于微服务拆分,将门店基础管理、会员体系、营销引擎、订单支付、多门店权限模块解耦。针对商家端,采用低代码可视化配置方案,商户无需编码,拖拽即可搭建小程序店铺、配置营销活动,大幅降低使用门槛。营销引擎作为核心模块,内置优惠券、储值、分销、拼团等常用算子,支持运营人员通过后台配置活动规则,无需后端迭代即可上线新玩法。

2508_94596546的博客 215

Isaac Lab 2.3 升级到 3.0:性能、多物理后端与 RSL-RL 的新变化

近期,把机器人强化学习的训练支持环境从 Isaac Lab 2.3.0 升级到了 3.0.0。最直观的感受是实际训练变快了,而进一步了解后,发现这次升级还带来了多物理后端、更灵活的学习库,以及更丰富的仿真与验证能力。本文围绕这些变化展开,接口迁移只简要带过。目前我仍主要使用 Isaac Sim/PhysX,尚未亲自测试其他物理后端;速度提升也暂未做严格量化。

CyanNoah的博客 294
上一篇: CS 336: Algorithms Problem
下一篇: normal OLS regression
wx____codinghelp
博客等级 码龄2年 771粉丝 115原创
评论
成就一亿技术人!
拼手气红包6.0元
还能输入1000个字符
 
 条评论被折叠 查看
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付:10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值