本課程首先介紹了機器學習、安全、隱私、對抗性機器學習和博弈論等主題。然后從研究的角度，討論各個課題和相關工作的新穎性和潛在的拓展性。通過一系列的閱讀和項目，學生將了解不同的機器學習算法，并分析它們的實現和安全漏洞，并培養開展相關主題的研究項目的能力。
//aisecure.github.io/TEACHING/2020_fall.html
Evasion Attacks Against Machine Learning Models (Against Classifiers) Evasion Attacks Against Machine Learning Models (Non-traditional Attacks) Evasion Attacks Against Machine Learning Models (Against Detectors/Generative odels/RL) Evasion Attacks Against Machine Learning Models (Blackbox Attacks) Detection Against Adversarial Attacks Defenses Against Adversarial Attacks (Empirical) Defenses Against Adversarial Attacks (Theoretic) Poisoning Attacks Against Machine Learning Models
付費5元查看完整內容
相關內容
Recent studies have shown that deep neural networks (DNNs) are highly vulnerable to adversarial attacks, including evasion and backdoor (poisoning) attacks. On the defense side, there have been intensive interests in both empirical and provable robustness against evasion attacks; however, provable robustness against backdoor attacks remains largely unexplored. In this paper, we focus on certifying robustness against backdoor attacks. To this end, we first provide a unified framework for robustness certification and show that it leads to a tight robustness condition for backdoor attacks. We then propose the first robust training process, RAB, to smooth the trained model and certify its robustness against backdoor attacks. Moreover, we evaluate the certified robustness of a family of "smoothed" models which are trained in a differentially private fashion, and show that they achieve better certified robustness bounds. In addition, we theoretically show that it is possible to train the robust smoothed models efficiently for simple models such as K-nearest neighbor classifiers, and we propose an exact smooth-training algorithm which eliminates the need to sample from a noise distribution. Empirically, we conduct comprehensive experiments for different machine learning (ML) models such as DNNs, differentially private DNNs, and K-NN models on MNIST, CIFAR-10 and ImageNet datasets (focusing on binary classifiers), and provide the first benchmark for certified robustness against backdoor attacks. In addition, we evaluate K-NN models on a spambase tabular dataset to demonstrate the advantages of the proposed exact algorithm. Both the theoretical analysis and the comprehensive benchmark on diverse ML models and datasets shed lights on further robust learning strategies against training time attacks or other general adversarial attacks.
//sites.google.com/view/ift6268-a2020/schedule
近年來，表示學習取得了很大的進展。大多數都是以所謂的自監督表示學習的形式。在本課程中，我們將對什么是自我監督的學習方法有一個相當廣泛的解釋，并在適當的時候包括一些無監督學習方法和監督學習方法。我們感興趣的方法，學習有意義的和有效的語義表示，而不(專門)依賴標簽數據。更具體地說，我們將對以下方法感興趣，如: 數據增廣任務，知識蒸餾，自蒸餾，迭代學習，對比方法 (DIM, CPC, MoCo, SimCLR等)，BYOL，以及自監督方法的分析。
我們的目標是了解自監督學習方法是如何工作的，以及起作用的基本原理是什么。
這是一個關于這一主題的高級研討會課程，因此，我們將閱讀和討論大量的最近的和經典的論文。講座將主要由學生主導。我們假設了解了機器學習的基礎知識 (特別是深度學習——正如你在IFT6135中看到的那樣)，我們還將探索自監督表示學習在廣泛領域的應用，包括自然語言處理、計算機視覺和強化學習。
在本課程中，我們將廣泛討論自監督學習(SSL)，特別是深度學習。最近，深度學習在許多應用領域取得了大量令人印象深刻的經驗收益，其中最引人注目的是在目標識別和圖像和語音識別的檢測領域。
在本課程中，我們將探討表示學習領域的最新進展。通過學生領導研討會，我們將回顧最近的文獻，并著眼于建立
本課程所涵蓋的特定主題包括以下內容:
- Engineering tasks for Computer Vision
- Contrastive learning methods
- Generative Methods
- Bootstrap Your Own Latents (BYoL)
- Self-distillation Methods
- Self-training / Pseudo-labeling Methods
- SSL for Natural Language Processing
- Iterated Learning / Emergence of Compositional Structure
- SSL for Video / Multi-modal data
- The role of noise in representation learning
- SSL for RL, control and planning
- Analysis of Self-Supervised Methods
- Theory of SSL
- Unsupervised Domain Adaptation
付費5元查看完整內容
本課程探索了生成式模型的各種現代技術。生成模型是一個活躍的研究領域: 我們在本課程中討論的大多數技術都是在過去10年發展起來的。本課程與當前的研究文獻緊密結合，并提供閱讀該領域最新發展的論文所需的背景。課程將集中于生成式建模技術的理論和數學基礎。作業將包括分析練習和計算練習。本課程專題旨在提供一個機會，讓你可以將這些想法應用到自己的研究中，或更深入地研究本課程所討論的主題之一。
- 自回歸模型 Autoregressive Model
- The NADE Framework
- RNN/LSTM and Transformers
- 變分自編碼器 Variational Autoencoders
- The Gaussian VAE
- ConvNets and ResNets
- Posterior Collapse
- Discrete VAEs
- 生成式對抗網絡 Generative Adversarial Nets
- f-GANs
- Wasserstein GANs
- Generative Sinkhorn Modeling
- 生成流 Generative Flow
- Autoregressive Flows
- Invertible Networks
- Neural Ordinary Differential Equations
- 基于能量的模型 Energy-Based Models
- Stein's Method and Score Matching
- Langevin Dynamics and Diffusions
付費5元查看完整內容
As data are increasingly being stored in different silos and societies becoming more aware of data privacy issues, the traditional centralized training of artificial intelligence (AI) models is facing efficiency and privacy challenges. Recently, federated learning (FL) has emerged as an alternative solution and continue to thrive in this new reality. Existing FL protocol design has been shown to be vulnerable to adversaries within or outside of the system, compromising data privacy and system robustness. Besides training powerful global models, it is of paramount importance to design FL systems that have privacy guarantees and are resistant to different types of adversaries. In this paper, we conduct the first comprehensive survey on this topic. Through a concise introduction to the concept of FL, and a unique taxonomy covering: 1) threat models; 2) poisoning attacks and defenses against robustness; 3) inference attacks and defenses against privacy, we provide an accessible review of this important topic. We highlight the intuitions, key techniques as well as fundamental assumptions adopted by various attacks and defenses. Finally, we discuss promising future research directions towards robust and privacy-preserving federated learning.
這是一門關于機器學習的入門課程。機器學習是一組技術，它允許機器從數據和經驗中學習，而不是要求人類手工指定所需的行為。在過去的20年里，機器學習技術在人工智能的學術領域和科技行業中都變得越來越重要。本課程提供了一些最常用的ML算法的廣泛介紹。
課程的前半部分側重于監督學習。我們從最近鄰、決策樹和集合開始。然后介紹了參數化模型，包括線性回歸、logistic回歸和softmax回歸以及神經網絡。然后我們轉向無監督學習，特別關注概率模型，以及主成分分析和k均值。最后，我們介紹了強化學習的基礎知識。
課程內容：
- 最近鄰導論
- 決策樹集成
- 線性回歸線性分類
- Softmax回歸、SVM、Boosting
- PCA、Kmeans、最大似然
- 概率圖模型
- 期望最大化
- 神經網絡
- 卷積神經網絡
- 強化學習
- 可微分隱私
- 算法公平性
//www.cs.toronto.edu/~huang/courses/csc2515_2020f/
推薦閱讀材料： Hastie, Tibshirani, and Friedman: “The Elements of Statistical Learning” Christopher Bishop: “Pattern Recognition and Machine Learning”, 2006. Kevin Murphy: “Machine Learning: a Probabilistic Perspective”, 2012. David Mackay: “Information Theory, Inference, and Learning Algorithms”, 2003. Shai Shalev-Shwartz & Shai Ben-David: “Understanding Machine Learning: From Theory to Algorithms”, 2014.
學習路線圖：
付費5元查看完整內容
強化學習理論(RL)，重點是樣本復雜性分析。
- Basics of MDPs and RL.
- Sample complexity analyses of tabular RL.
- Policy Gradient.
- Off-policy evaluation.
- State abstraction theory.
- Sample complexity analyses of approximate dynamic programming.
- PAC exploration theory (tabular).
- PAC exploration theory (function approximation).
- Partial observability and dynamical system modeling.
付費5元查看完整內容
生成式對抗網絡
·
【導讀】Pieter Abbeel 是加州大學伯克利分校的教授，伯克利機器人學習實驗室的主任，其新開課程CS294深度無監督學習包含兩個領域，分別是生成模型和自監督學習。這個15周的課程包含視頻PPT能資源，有助于讀者對深度學習無監督的理解。最新一期是生成式對抗網絡Generative Adversarial Networks的課程，共有257頁ppt，包括GAN, DC GAN, ImprovedGAN, WGAN, WGAN-GP, Progr.GAN, SN-GAN, SAGAN, BigGAN(-Deep), StyleGAN-v1,2, VIB-GAN, GANs as Energy Models，非常值得關注！
目錄內容：
- 隱式模型的動機和定義
- 原始GAN (Goodfellow et al, 2014)
- 評估: Parzen、Inception、Frechet
- 一些理論: 貝葉斯最優鑒別器; Jensen-Shannon散度; 模式崩潰; 避免飽和
- GAN進展
- DC GAN (Radford et al, 2016)
- 改進GANs訓練(Salimans et al, 2016)
- WGAN, WGAN- gp, Progressive GAN, SN-GAN, SAGAN
- BigGAN, BigGAN- deep, StyleGAN, StyleGAN-v2, VIB-GAN
- 創意條件GAN
- GANs與申述
- GANs作為能量模型
- GANs與最優傳輸，隱式似然模型，矩匹配
- 對抗性損失的其他用途:轉移學習、公平
- GANs和模仿學習
付費5元查看完整內容
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191