Next steps · after page 14
Further learning
The demos in this series run in your browser. The next step is writing real code on real data. Start with a free notebook in Google Colab, then use the courses, books, and datasets below to keep going.
Run your first model in five minutes
1.Go to colab.research.google.com and sign in with a Google account.
2.Choose New notebook. NumPy, pandas, scikit-learn, matplotlib, and PyTorch are already installed.
3.Paste the code below into a cell and press Shift + Enter.
4.Change the model, the features, or the threshold, and rerun. Compare with what you saw on pages 06 to 09.
from sklearn.datasets import load_breast_cancer from sklearn.model_selection import train_test_split from sklearn.preprocessing import StandardScaler from sklearn.linear_model import LogisticRegression from sklearn.metrics import classification_report X, y = load_breast_cancer(return_X_y=True) # a built-in dataset: 569 rows, 30 features X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, stratify=y, random_state=0) scaler = StandardScaler().fit(X_train) # page 05 model = LogisticRegression(C=1.0) # page 08, C = 1/lambda from page 07 model.fit(scaler.transform(X_train), y_train) y_pred = model.predict(scaler.transform(X_test)) print(classification_report(y_test, y_pred)) # page 09
Prefer working locally? Install Python from python.org, then follow the virtual environment steps on page 03.
Courses and videos
COURSEGoogle Machine Learning Crash CourseShort, free lessons with interactive exercises. A good next step right after this series.COURSEMachine Learning SpecializationAndrew Ng's introductory series on Coursera. Free to audit; covers regression, classification, neural networks, and clustering.COURSEPractical Deep Learning for Codersfast.ai's free, code-first course. You train real image and text models from the first lesson.PRACTICEKaggle LearnFree bite-size notebooks on Python, pandas, machine learning, and deep learning, run in the browser.VIDEO3Blue1Brown: Neural networksBeautifully animated videos on what neural networks, gradient descent, backpropagation, and transformers are actually doing.VIDEOStatQuest with Josh StarmerClear, friendly YouTube explanations of nearly every algorithm and metric in this series.
Books and documentation
BOOKAn Introduction to Statistical LearningThe classic textbook, with a Python edition (ISLP). Free PDF from the authors' site.BOOKDive into Deep LearningFree interactive book where every chapter is runnable code, in PyTorch and other frameworks.DOCSscikit-learn User GuideThe official guide. Each model page explains the math and shows examples.DOCSPyTorch TutorialsOfficial step-by-step tutorials, from tensors to training CNNs and transformers.
Datasets to practice on
DATAUCI Machine Learning RepositoryHundreds of classic, well-documented datasets for every kind of problem.DATACIC-IDS2017Labeled network traffic with modern attacks, from the Canadian Institute for Cybersecurity. Good for intrusion detection projects.DATANSL-KDDA cleaned-up version of the classic KDD Cup 1999 intrusion dataset. Small and easy to start with, though dated.DATAEMBERFeatures extracted from over a million Windows executables, labeled malicious or benign, from Elastic.
Handle security datasets with care. Some contain real malware samples or captured traffic. Work in an isolated lab environment, never on a production or personal machine, and follow each dataset's license.
AI security frameworks
FRAMEWORKMITRE ATLASA knowledge base of real attacks on AI systems, organized like MITRE ATT&CK, with case studies.FRAMEWORKOWASP Top 10 for LLM ApplicationsThe most critical risks in apps built on large language models, including prompt injection.FRAMEWORKOWASP Machine Learning Security Top 10Top risks for traditional ML systems, such as poisoning, evasion, and model theft.FRAMEWORKNIST AI Risk Management FrameworkVoluntary guidance for governing, mapping, measuring, and managing AI risk in an organization.