/ece364_2023

Primary LanguageJupyter NotebookBSD 3-Clause "New" or "Revised" LicenseBSD-3-Clause

ECE 364: Machine Learning for Predictive Data Analytics

License Python 3.7, 3.8 Hits

Welcome to ECE 364! In this course we present basic concepts in machine learning for predictive data analytics:

  • Information-based learning: decision trees; Shannon’s entropy model; information gain; ID3 algorithm; feature selection; predicting continuous targets; tree pruning; model ensembles: boosting, bagging.
  • Similarity-based learning: feature spac, distance metrics, nearest neighbor algorithm, noisy data, predicting continuous targets, various measures of similarity, feature selection.
  • Probability-based learning: Bayes' theorem; Bayesian prediction; curse of dimensionality; conditional independence and factorization; Naive Bayes model; smoothing; continuous features: probability density functions, binning; Bayesian networks.
  • Error-based learning: simple linear regression; error surface; multivariate linear regression; gradient descent; learning rate; handling categorical features; modeling nonlinear relationships; multinomial logistic regression; support vector machines.
  • Deep learning: artificial neural networks; activation functions; backpropagation and gradient descent; vanishing gradients; weight initialization; categorical target features: softmax layer and cross-entropy loss; dropout.
  • Evaluation: misclassification rate; confusion matrix: precision, recall, F1 measure, profit/loss; prediction scores: receiver operating characteristic curve, Kolmogorov-Smirnov statistic, gain/lift; measures of error; evaluating models after deployment.
  • The art of machine learning: perspectives on prediction models: parametric vs. nonparametric, generative vs. discriminative; choosing a machine learning approach: no free lunch theorem, project requirements, data considerations.

Useful Links

  • Canvas web site
  • Zoom links and TA contact information can be found on Canvas.

Course staff

  • Instructor: Niraj K. Jha
  • TAs: Sayeri Lala, Margarita Belova

Timings

  • Lectures: M/W 3:00-4:20pm (EQuad B205)
  • Office hours:
    • Niraj K. Jha: M/W 2-3pm (EQuad B205)
    • Sayeri Lala: Tu: 4-5pm, Th: 11-12pm (EQuad B321)
    • Margarita Belova: M: 11-12pm, W:1-2pm (EQuad B321)

Grading

  • Assignments (25%):
  • Mid-term exam (25%)
  • Final exam (50%)

Reading

License

BSD-3-Clause. Copyright (c) 2023, JHA-Lab. All rights reserved.

See License file for more details.