Whether you're launching a career in data science, developing expertise in deep learning and AI, or preparing for your next opportunity, WQU Labs provide a flexible way to build practical, in-demand technical skills through hands-on projects.
These self-paced, project-based programs are designed as applied credentials, giving learners the opportunity to develop and demonstrate their skills through authentic project work using real-world datasets. Working in cloud-based virtual machines, learners tackle the kinds of challenges faced by data scientists and AI practitioners across industries - entirely free of cost.
Self-Paced (Est. 10-16 weeks)
10-15 Hours Per Week
Deep Learning Fundamentals Lab:
The Deep Learning Fundamentals Lab bridges the gap between data science foundations and AI mastery. Across two progressive units of six projects each, you'll design and train deep learning models, implement CNNs with PyTorch, and work with real datasets spanning health, science, and engineering. Unit 1 earns you a shareable digital badge, with Unit 2 leading to full Lab certification and preparing you for specialization in Computer Vision, NLP, and LLMs.
Prerequisites:
Intermediate-level Python skills
Basic calculus & linear algebra
Basic probability & statistics
Experience with data science concepts
Recommended machine learning experience
Application Requirements
Above Suggested Prerequisites
Passing score on Admissions Assessment (70% or higher)
The Deep Learning Fundamentals Lab is an advanced learning opportunity designed to help you master the core concepts behind deep learning through two units of six hands-on projects each - ranging from using PyTorch to build models to applying CNNs to real-world problems. Applicants are expected to have the following prerequisite skills:
Basic linear algebra (i.e., matrices, vectors, and matrix operations)
Basic calculus concepts (i.e., function analysis, derivatives, gradients, etc.)
Basic probability and statistics functions
Intermediate-level Python programming, including: basic data structures like arrays and dictionaries, the ability to write definitions for functions and classes, and familiarity with data manipulation using libraries like NumPy and Pandas.
Familiarity with essential machine learning concepts, including supervised and unsupervised learning, overfitting and regularization, and training, validation, and test sets
Before you attempt the Admissions Assessment, we recommend that you use the following free resources to help you prepare:
Mathematics for Machine Learning: Comprehensive coverage of linear algebra, calculus, and probability.
Algebra Basics at Khan Academy: Foundations, algebraic expressions, linear equations and inequalities, graphing lines and slope
Statistics and Probability at Khan Academy: Displaying and comparing quantitative data, summarizing quantitative data, modeling data distributions.
Python at LearnPython.org: Learn the Basics
Applied Data Science Lab: WQU’s own Applied Data Science Lab is free and always available. The Applied Data Science Lab teaches you the Python and Machine Learning skills needed to succeed in the Deep Learning Fundamentals Lab.
If you fail the Admissions Assessment
You have a second (and final) attempt after a 7-day waiting period.
Applicants who do not pass the test on their 2nd attempt are able to reapply to the Lab following a waiting period of 6 months from the date of their 2nd attempt.
Important Warning: Creating multiple accounts to attempt the Admissions Assessment is a violation of the University’s Academic Integrity policy. Any user identified for doing so will be immediately terminated and will not have the opportunity to be considered for enrollment in the Lab program.
Next Steps After Passing the Admissions Assessment
Complete your Student Profile;
Sign the User Agreement;
Take the mandatory Orientation Course;
Register for Project 1.
Once you have completed your profile and signed the User Agreement, you will be automatically enrolled in a mandatory Orientation Course, which can be accessed via My Courses in the top navigation of the WQU Learning Platform.
From walking you through the Lab’s structure to helping you navigate our Learning Platform and virtual machines, the Orientation Course takes about 1 hour to complete and covers everything you need to know to set you up for success.
Upon successful completion of the Orientation Course, course registration will NOT occur automatically. Instead, you are responsible to register for each of your Projects by navigating to My Courses and "Register".
By the end of this lab, you’ll know how to:
Build and evaluate deep neural networks
Apply CNNs to solve real-world problems
Use PyTorch to train models from scratch
Optimize performance using transfer learning, regularization, and more
Credentials Earned Upon Successful Completion
Verified Digital Badge and Certificate
Project Descriptions
The Deep Learning Fundamentals Lab comprises two units, each with six end-to-end projects. Each successful project completion unlocks the registration for the next.
Predicting Concrete Strength with Deep Learning Regression
Project 1 • self-paced
This project introduces learners to the fundamental building blocks of deep learning using PyTorch. Learners begin by manipulating tensors and understanding real-world data representations. From there, they progressively build models, starting with linear regression and culminating in a simple multi-layer neural network (MLP). Using the Concrete Compressive Strength dataset, learners explore the relationship between model complexity, non-linearity, and prediction performance.
Diagnosing Heart Disease with Neural Networks
Project 2 • self-paced
Building on the foundation laid in Project 1, this project guides learners through the core concepts and implementation of simple neural networks. Beginning with the perceptron algorithm, learners explore how neurons form networks, implement both forward and backward passes manually, and then use PyTorch to train and evaluate their models. The Heart Disease dataset provides practice with binary classification and helps learners understand the impact of model structure, non-linearity, and optimization.
Classifying Protein Localization Sites Using Deep Learning
Project 3 • self-paced
This project introduces learners to multiclass classification using the Yeast dataset, a bioinformatics dataset involving 10 classes. Through four structured notebooks, learners explore new loss functions, optimization techniques, and evaluation metrics. They also investigate strategies for improving generalization, such as regularization, while tackling issues like class imbalance.
Training Neural Architectures for Image Classification
Project 4 • self-paced
Building on multi-class classification, this project introduces deep neural networks using the CIFAR-10 image dataset. Learners explore the training challenges of deeper architectures, including initialization, optimization strategies, and stabilization techniques. The project concludes with simplified implementations of LeNet and AlexNet to transition into CNNs in the next project.
Pet Breed Classification with CNNs
Project 5 • self-paced
This project marks the transition from fully connected networks to convolutional architectures, equipping students with the core tools for computer vision tasks. Using the Oxford-IIIT Pet Dataset, learners explore how convolutional layers extract spatial patterns and textures from image data. Learners implement, train, and evaluate their first CNNs, and reflect on their performance relative to deep MLPs. The project concludes with a structural introduction to foundational CNNs like LeNet and VGG, preparing the ground for advanced vision architectures in the next module.
Advanced Vision Models for Seedling Classification
Project 6 • self-paced
This project deepens learners' experience with convolutional architectures by exploring Transfer Learning and Advanced CNNs. Using the Plant Seedlings Classification dataset, learners work with real-world images of plants to build resilient models in the face of visual variability like lighting, angles, and background noise. The project focuses on modern strategies for improving training efficiency and model generalization, including fine-tuning pre-trained models and applying data augmentation techniques. This transition prepares learners to tackle production-ready image classification tasks using cutting-edge deep learning techniques.
From Sensors to Actions: RNN-Based Human Activity Recognition
Project 7 • self-paced
This project introduces the fundamentals of Recurrent Neural Networks (RNNs) and how they process sequential data. Learners work with the UCI Human Activity Recognition (HAR) Dataset, a multivariate time series collected from smartphone sensors. The task is to classify human activities such as walking, standing, or laying based on sensor signals. Learners build simple RNNs from scratch in PyTorch, analyze hidden state behavior, and evaluate performance, while also understanding challenges like the vanishing gradient problem.
Deep Learning for Weather Forecasting with LSTMs
Project 8 • self-paced
This project explores the design and application of Long Short-Term Memory (LSTM) networks for time series forecasting. Learners use the Global Surface Temperature Time Series Dataset, a univariate sequence of daily minimum temperatures. They begin with framing time series into supervised learning problems, then implement LSTMs and GRUs in PyTorch to capture long-term dependencies. The project emphasizes sequence-to-sequence learning, multi-step forecasting, and comparing LSTM vs. GRU architectures.
The Art of Compression, Autoencoders and Variational Models
Project 9 • self-paced
This project introduces students to autoencoders and variational autoencoders (VAEs), two foundational architectures in the field of generative deep learning. Using a curated subset of the FER2013 facial expression dataset, learners explore how neural networks can compress, reconstruct, and generate images by learning meaningful latent representations. The project covers practical applications such as dimensionality reduction, anomaly detection, and image denoising, before transitioning to probabilistic generative modeling with VAEs. Through visualization of latent spaces and reconstructed faces, students develop both technical understanding and intuition about how deep networks capture underlying data structures.
Adversarial Creativity Generating Art with GANs
Project 10 • self-paced
This project immerses students in the world of Generative Adversarial Networks (GANs), one of the most revolutionary architectures in modern generative modeling. Using a curated subset of the WikiArt dataset, learners explore how GANs can synthesize new artworks by learning from diverse artistic styles. Through a sequence of hands-on notebooks, learners move from understanding the Generator–Discriminator interplay to implementing simple GANs and deep convolutional variants (DCGANs). They then use pre-trained models to generate visually compelling paintings, examine common challenges like mode collapse and instability, and experiment with creative latent-space manipulations. The project concludes with a critical reflection on the artistic potential and ethical implications of AI-generated imagery.
Transferring Knowledge Fine Tuning on Caltech Images
Project 11 • self-paced
This project introduces students to the practical power of transfer learning, a technique that enables deep neural networks to reuse knowledge acquired from large scale datasets. Using a curated subset of Caltech 101, learners explore how pre-trained convolutional neural networks such as ResNet or VGG models originally trained on ImageNet can be adapted to classify new image categories efficiently and with minimal data.
Across a sequence of hands on notebooks, students begin by performing feature extraction, freezing pre-trained layers and training a lightweight classifier tailored to the Caltech classes. They then experiment with fine tuning, selectively unfreezing deeper layers to improve specialization and overall performance. Along the way, learners visualize predictions, compare both strategies, and build intuition for why transfer learning is one of the most impactful techniques in modern computer vision.
The project concludes with a broader reflection on the importance of reusing pre-trained knowledge, highlighting how transfer learning accelerates experimentation, reduces computational cost, and delivers strong results even in low data scenarios.
Exploring Responsible and Sustainable AI
Project 12 • self-paced
This project concludes the deep learning sequence by inviting learners to explore the ethical societal and environmental dimensions of modern AI systems. Instead of training new models, learners work through a series of visual case studies and guided reflections that illustrate how data collection model design and deployment choices influence fairness, accountability, trust and sustainability.
Using carefully curated image examples and lightweight analytical exercises students examine real world scenarios involving dataset imbalance bias in computer vision systems synthetic media and deepfake creation and the growing energy cost of training large scale neural networks. Each notebook blends narrative explanation with simple visualizations encouraging students to connect their technical understanding with broader questions about responsibility and impact.
The project culminates in a forward looking discussion of emerging trends including foundation models, multimodal learning and new hardware paradigms while emphasizing the importance of developing AI that is inclusive, transparent and environmentally conscious.
Learning Outcomes
Synthesize Core Deep Learning Concepts
Synthesize core deep learning concepts, architectures, and mathematical foundations to explain how neural networks process information and learn from data.
Identify Deep Learning Applications Across Domains
Identify specific deep learning applications in healthcare, computer vision, and reinforcement learning domains, and determine whether CNN, feedforward networks, or advanced architectures are most suitable for classification, regression, or pattern recognition tasks.
Execute Complete Neural Network Training Workflows
Execute complete neural network training workflows, including data preparation, model configuration, backpropagation implementation, and performance optimization.
Construct and Modify CNN Architectures
Construct and modify CNN architectures from basic networks to advanced models (ResNet, Inception), implementing transfer learning and data augmentation techniques.
Diagnose and Optimize Training Challenges
Diagnose training challenges (overfitting, gradient problems), evaluate optimization strategies, and select appropriate regularization and architectural solutions.
Measure and Compare Model Performance
Measure model performance using accuracy, precision, recall, and loss metrics; compare computational efficiency between LeNet, AlexNet, and VGG architectures; and optimize learning rates, batch sizes, and dropout parameters to achieve target performance benchmarks.
Implement Smart Imputation Strategies
Apply context-aware imputation including group-based imputation, interpolation methods, and domain-specific rules while understanding their impact.
Design Robust Data Filters
Create multi-level filtering strategies, handle edge cases, and validate filtering decisions through sensitivity analysis.
Extract Complex Features
Engineer meaningful features from raw data, create interaction variables, and design domain-specific transformations.
Assess Data Quality Systematically
Build comprehensive data quality frameworks, create validation rules, and quantify the reliability of datasets.
Perform Sensitivity Analysis
Test how different data preparation choices affect results and document the impact of various cleaning decisions.
Handle Complex Missing Patterns
Identify and address different types of missingness (MCAR, MAR, MNAR) using appropriate strategies.
Create Reproducible Pipelines
Build modular, well-documented data preparation workflows that can be validated and modified.
Prepare Analysis-Ready Datasets
Output datasets optimized for different analytical purposes with clear documentation of all transformations.