一個簡單的操作手冊，用於hack the box。

HACK THE BOX 操作手冊

這個系列文章不會講授太多的數學原理，想看數學可以看參考文章，但是想從代碼層面上細講，因為這比推算一個公式簡單多了

從零開始實現 Stable Diffusion - 1

CodeDaily 是一個鼓勵每日提交一份自己的題解上倉庫，不過項目到了最後大部分都是我的提交，可能想靠leetcode刷題的人少之又少吧

2972. 统计移除递增子数组的数目 II

+10x 工程師可能是虛構的，但 -10x 工程師是存在的。要成為 -10x 工程師，每週只需浪費 400 個工程時間。結合以下策略：

如何成为 -10x 工程师 How to be a -10x Engineer

之前一直在Unity搞卡通渲染，3D場景製作完全沒有接觸過。在虛幻5用到輔助工具比Unity真的好太多太方便了，在工具和方便程度來講，UE5確實真的比得上Unity的。在GameJam上之前多數用的Unity，Unity在製作2D、定製功能比UE多的多，但更考驗遊戲製作者的架構設計能力，做不好的話會導致遊戲功能太多繁瑣。

如何在虛幻5製作原神地形教程

At the heart of the decision to use minibatches is computational efficiency. This is most easily understood when considering parallelization to multiple GPUs and multiple servers. In this case we need to send at least one image to each GPU. With 8 GPUs per server and 16 servers we already arrive at a minibatch size no smaller than 128.

Deep learning Guide 14: Minibatch Stochastic Gradient Descent

Although it is rarely used directly in deep learning, an understanding of gradient descent is key to understanding stochastic gradient descent algorithms.

Deep learning Guide 12: Gradient Descent 梯度下降

我们一直在训练过程中使用随机梯度下降，但没有解释它为什么起作用。为了澄清这一点，我们刚在 11.3节中描述了梯度下降的基本原则。本节继续更详细地说明随机梯度下降（stochastic gradient descent）。

Deep learning Guide 13: Stochastic Gradient Descent 隨機梯度下降

Assume that we want to verify convexity of a set by drawing all lines between points within the set and checking whether the lines are contained. 

課後功課答案  12.2. Convexity

地表最強のTransformer Architecture: layout normal + residual connection + 

Deep learning Guide 11: The Transformer Architecture

Implement distance-based attention by modifying the DotProductAttention code. Note that you only need the squared norms of the keys for an efficient implementation. 

課後功課答案 11.5. Self-Attention and Positional Encoding

課後功課答案 11.4. Multi-Head Attention

In deep learning, we often use CNNs or RNNs to encode sequences. Now with attention mechanisms in mind, imagine feeding a sequence of tokens into an attention mechanism such that at every step,

Deep learning Guide 10: Multi-Head Attention, Self-Attention and Positional Encoding

課後功課答案 11.3. Attention Scoring Functions

mack1982weak證明了Nadaraya-Watson估計是一致的。當您獲得更多數據時，您應該以多快的速度縮小注意力機制的規模？為你的答案提供一些直覺。

課後功課答案 11.2. Attention Pooling by Similarity

Suppose that you wanted to reimplement approximate (key, query) matches as used in classical databases, which attention function would you pick?

課後功課答案 11.1. Queries, Keys, and Values

來到deep learning的終點: Attention Mechanisms

Deep learning Guide 9: Attention Mechanisms

Despite having just one hidden layer between the input at any time step and the corresponding output, there is a sense in which these networks are deep

Deep learning Guide 8: 深度循环神经网络、双向循环神经网络

LSTMs are the prototypical latent variable autoregressive model with nontrivial state control. Many variants thereof have been proposed over the years

Deep learning Guide 7: Long Short-Term Memory (LSTM), Gated Recurrent Units (GRU)

Recurrent neural networks (RNNs) are neural networks with hidden states. Before introducing the RNN model,

Deep learning Guide 6: Recurrent Neural Networks

A distribution over sequences satisfies the Markov property of first order

Deep learning Guide 5: Sequence Models, Language Models

因為想在 Stable Diffusion WebUI 消除圖片的Metadata，但又不能單獨移除，很煩

Stable Diffusion WebUI Extension 開發入門 

Training deep neural networks is difficult. Getting them to converge in a reasonable amount of time can be tricky. In this section, we describe batch normalization, a popular and effective technique

Deep learning Guide 4: Batch Normalization, ResNet, DenseNet, 面試經典題

one simple strategy: (i) use 1×1 convolutions to add local nonlinearities across the channel activations and (ii) use global average pooling to integrate across all locations in the last representation layer

Deep learning Guide 3: VGG, NiN, GoogLeNet

Deep Convolutional Neural Networks (AlexNet) and Representation Learning

Deep learning Guide 2: LeNet model

Deep learning Guide 1: Layers and Modules 

Abalone from the UCI Machine Learning Repository and inspect their properties.

Tools for Deep Learning 1: speeding up your workflow 

Download that creepy costly dump plugin from marketplace.

How to specify a default Jupyter kernel via configuration?

SQLlite 安裝和教學

Warren McCulloch and Walter Pitts in 1943. They proposed a model of an artificial neural network anddemonstrated that simple network structures could learnMcCulloch (second) created the corner stone of neural computing and artificial neural networks (ANN). (人工神經網路)

Machine Learning Note: (Translation)

Angular網站開發: 從表單開始

紀錄一下我是如何從零開始製作一個三渲二人物的

Unity 崩壞: 星穹鐵道人物三渲二渲染器製作

托福流水帳2

倉庫管理會內的Leetcode和其他刷題網站的代碼分析文章，但裡面總有出現代碼錯誤和格式沒寫全的問題，這裡使用Github Action進行全方面檢查

使用 Github Action 檢查文章格式和對代碼進行單元測試

使用FFmpeg轉換多媒體文件

之前在更新LeetCode表單內容的時候，一直需要手動填寫leetcode編號和名字，非常煩人。

使用GitHub Action全自動更新倉庫

渲染相關方程、HLSL程式筆記

講述如何使用Blender、模之屋、mmd插件、最後加入到虛幻5場景中

如何把原神角色加入場景虛幻5教程

遊戲槍械的後座力研究(上鎖)

澳門大學2025QS排名分數構成分析

寫了個自動執行SQL腳本的化碼，帶輸出日誌

Windows SQL自動執行腳本

記一下哪裡寫錯了密碼: “sTEPhenLaM”

Toefl Official 學習筆記

Leetcode 2813. 子序列最大优雅度

ShaderToy是一個線上工具，可以讓你用Cg或GLSL來寫Shader，然後在線上跑看看效果。

友情連結

tom-ci

交流頻道

加入我們的社群討論分享