Deep Learning Engineering : Applied Projects

Explore essential math symbols for deep learning, including limits, derivatives, integrals, gradients, matrices, transforms, probability, and KL divergence.

Define mu as the mean, sigma as standard deviation, and variance; illustrate covariance and correlation rho; discuss supremum and infimum. Also cover element of, and real numbers R and R^n.

Explore math symbols from integers Z and natural numbers N to complex C and rational Q. Apply logical operators, summation, and parameters like lambda and alpha in optimization and learning.

Explore Python, a high level interpreted language celebrated for readability and versatility. See how its standard library, third-party packages, and multi-paradigm support drive web development, data science, AI, and automation.

Explore how Anaconda, Jupyter, and VSCode together streamline Python development by managing environments, running interactive notebooks, and enabling code, debugging, and version control.

Explore Google Colab, a browser-based cloud platform for Python, data analysis, and machine learning with TensorFlow, Keras, NumPy, and SciPy pre-installed, GPU access, Google Drive integration, and real-time collaboration.

Install the latest stable Python, Anaconda, and Visual Studio Code, verify installations via the command line, and enable path options, including the Python extension in VS Code.

Explore Python syntax, variables, and data types, including integers, floats, strings, booleans, and none, and master type conversions, dynamic typing, and basic arithmetic, comparison, and logical operations with precedence rules.

Explore lists, tuples, and sets in Python, learn mutability and immutability, indexing and slicing, and apply operations like append, insert, remove, and union, intersection, difference.

Master Python control structures with conditional statements such as if, elif, else, and loops, using range, nesting, and logical operators, break and continue to manage flow and prevent infinite loops.

Explore functions and basic functional programming in Python, including def, parameters, return values, and higher‑order constructs with lambda, map, filter, and reduce for data manipulation.

Explore intermediate functions including recursion, tail recursion, currying, partial functions, closures with state, decorators, and generators, and review map, filter, reduce for modular, reusable, and lazy Python code.

Learn how Python dictionaries store data as unique key-value pairs, access and modify entries with brackets or get, and use comprehensions to transform, filter, and swap mappings.

Explore Python modules, packages, and import techniques, and learn to use the standard library with NumPy and pandas for data analysis, DataFrame manipulation, and missing data handling.

Explore Python file handling to read, write, and close files across text, csv, and structured formats, using open, read, write, and close, plus csv and pandas workflows.

Master exception handling and robust code in Python by using try, except, finally, and custom exceptions to gracefully handle errors like division by zero and missing files.

Learn object oriented programming in Python, focusing on classes and objects, encapsulation, inheritance, and polymorphism, with practical concepts like attributes, methods, constructors, and multiple inheritance.

Master data visualization in Python using Matplotlib and Seaborn, creating line plots, bar charts, histograms, heatmaps, and kernel density estimates for clear, publication-ready insights.

Explore advanced list operations in Python, including list comprehensions, enumerate and zip, with filtering, nested loops, and transforming data for concise, readable code.

Explore how neural networks learn by processing data through neurons with weighted connections. Understand activation functions like sigmoid, ReLU, and tanh, plus forward and backward propagation, loss, and learning rate.

Explore vectors, their magnitude, direction, components, and key operations—addition, subtraction, scalar multiplication, and dot product—and their role in data representation, gradients, and minimizing loss in machine learning.

Explore eigenvalues and eigenvectors of matrices, solve the characteristic equation, and apply these concepts to PCA, quantum mechanics, control systems, and Markov chains.

Explore probability theory foundations, including experiments, sample spaces, events, and probability rules, and learn about discrete and continuous distributions, Bayes' theorem, and random variables.

Explore how functions map inputs to outputs, covering domain, range, injective, surjective, and bijective properties, with linear to trigonometric examples for feature transformation and activation in machine learning.

Explore derivatives as rates of change and gradient concepts, covering first through third derivatives, backpropagation and gradient descent, activation functions, and rules—power, constant, product, quotient, chain—with SymPy Python code.

Learn how to ensure data quality through profiling, cleaning, validation, and automation to produce accurate, complete, consistent, timely, and relevant data for reliable analyses.

Clean data enhances accuracy, reliability, and usability by addressing missing values, duplicates, outliers, invalid data, and inconsistencies, using imputation, deduplication, standardization, validation, and pipelines.

Learn data pre-processing strategies for missing values, including MCAR, MAR, and MNAR, with deletion, mean/median/mode imputation, forward/backward filling, interpolation, and KNN or multivariate imputation.

Identify outliers—global, contextual, and collective—using boxplots, histograms, and scatter plots; apply z-score or IQR, and model-based approaches like isolation forest and DBSCAN.

Standardization centers data at zero and scales to unit variance to ensure equal feature contribution and model convergence. It includes z-score, robust, and group-wise methods, applying to PCA and SVM.

Encode categorical variables into numeric representations to power machine learning models. Learn label, one-hot, binary, ordinal, target, and frequency encodings and their nominal vs ordinal use.

Master feature engineering to transform data into features that boost model performance. Apply understanding, exploratory data analysis, and feature creation, transformation, and selection across text, time series, and structured data.

Reduce dimensionality by lowering feature count to cut costs and overfitting, preserving information for visualization, using feature selection and extraction methods like PCA, LDA, t-SNE, and autoencoders.

Explore gradient descent, the optimization method that minimizes a cost or loss function by iteratively updating model parameters via the gradient and learning rate.

Build and train a deep feedforward neural network to learn XOR using the Adam optimizer, various activations, and binary cross-entropy, validating predictions on XOR data.

Build a deep feedforward network with three hidden layers to classify MNIST digits, loading and preprocessing data, flattening to 784 inputs, and training with batch normalization, dropout, softmax, and Adam.

design and implement the mlp mixer in pytorch, building an mlp block and token and channel mixer to mix across patches and channels using layernorm, gelu, and dropout.

Build a PyTorch cnn by defining a network with conv layers, relu, and max pooling to extract features from a 3-channel input, then flatten and map to 10 outputs.

Build a cnn in Julia with Flux, using two conv layers (3x3, 16 then 32 channels), padding, relu, 2x2 max pooling, flattening, and two dense layers for classification.

Explore a MATLAB cnn architecture that processes 32x32x3 images through two conv-pool blocks, followed by fully connected layers, softmax, and classification for ten classes.

Watch Yann LeCun's 1993 demo of one of the first convolutional neural networks for text recognition, processing handwritten digits in real time and laying the foundation for OCR.