DS212BKK
Intro to Deep Learning
Faculty
Mikhail Romanov
Senior Machine Learning Engineer, Yandex, Expert
Course length
Duration
Total hours
Credits
Language
Course type
Fee for single course
Fee for degree students
Skills you’ll learn
Overview
By completing this course, students will gain fundamental knowledge and practical skills in deep learning, their first step towards becoming a data scientist.
To start, we'll go through the basics of neural networks. We will explore their architecture and tuning algorithms, gaining a deep understanding of everything that happens after clicking "Start Training". We will discuss how to best present a problem to a neural network since not all problems are solvable in principle, and in this, the method of maximum likelihood will assist us.
Next, we will explore two major application areas of neural networks: computer vision and natural language processing. To gain a good understanding of the former, you will study convolutional neural networks, regularisation methods, and normalisation. As for natural language processing, we will discuss transformers, BERT, and GPT. We will also learn how image and text representations can be placed in the same space using CLIP.
Learning highlights
- Ability to formulate a problem in terms of machine learning.
- Knowledge of specific machine learning tasks such as regression and classification, detection, segmentation, text, and image generation
- Knowledge of deep learning architectures: dense and convolutional neural networks, transformers, and their variations
- Ability to train a deep learning model for a specific business task.
- Knowledge of basic metrics for evaluating the quality of models.
Course outline
15 classes
Neuron and Neural Network
- Mathematical Model of a Neuron
- Theoretical Problems
- Boolean Operations as Neurons
- From Neuron to Neural Network
- Practice: Basic Work in PyTorch
- Building your First Neural Network
Building your First Neural Network:
- Recovering Dependencies with Neural Networks
- Components of a Neural Network
- Neural Network Tuning Algorithm
- Theoretical Problems: Computational Graphs and BackProp
- Theoretical Problems: Dependency Recovery
- Practice: Implementing Gradient Descent
Tasks Solved with Neural Networks in Computer Vision
- Binary Classification? Binary Cross-Entropy!
- Multiclass Classification? Softmax!
- Localization, Detection, Segmentation, and Super-Resolution
- Theoretical Problems: Loss Functions
- Practice: Building Your First Neural Network
- Practice: Classification in PyTorch
Optimisation Methods
- Basic Gradient Descent
- Modifications of Gradient Descent
- Theoretical Problems: Understanding SGD with Momentum
- Practice: Classification of Handwritten Digits with a Fully Connected Network
Convolutional Neural Networks
- Convolution, Convolution Cascades
- Practice: Implementing a Convolutional Layer
- Collect Them All: LeNet Architecture (1998)
- Collect Them All: AlexNet (2012) and VGG (2014)
- Collect Them All: GoogLeNet and ResNet (2015)
- Practice: Recognizing Handwritten Digits with a Convolutional Neural Network
- Theoretical Problems: Architectures of Convolutional Neural Networks
Regularization and Normalisation
- Fighting Overfitting with Dropout!
- Not Enough with Eternity? Batch Normalisation!
- Practice: Normalisation Layer
- Practice: Solving a Classification Problem on the CIFAR Dataset
- Theoretical Problems: Regularisation
Maximum Likelihood Method and Competition
- Maximum Likelihood Method
- Theoretical Problems: Maximum Likelihood Method
- Practice: Transfer Learning with a Kaggle Competition Example
Fundamentals of Natural Language Processing
- Text Embeddings
- Practice: Text Classification through One-Hot Encoding and a Linear Model
Text Embeddings
- Unsupervised Learning
- Word2Vec and FastText
- Practice: Emotion Classification Competition for Twitter Posts
Transformer Architecture
- Positional Encodings
- Text Translation Task from One Language to Another
- Attention Mechanism and the Quadratic Time Computation Problem
Splitting the Transformer into Two Parts
- BERT and GPT
- HuggingFace Library
- Practice: Competition on Determining Programming Language from Code
Transformers for Computer Vision
- Discussing the Paper "An Image is Worth 16x16 Words"
- Practice: Image Classification Task in the CIFAR Dataset through Vision Transformer
Combining Images and Text
- CLIP Architecture
Recap of the Course
- Recap of the Course
Final Exam
Final Exam
Course materials
Media
Prerequisites
Python
Students need to have basic knowledge of linear algebra and calculus. They must remember what the equation for the plane looks like and what the “gradient” is.
Methodology
Each lesson lasts 3 hours. We study new material and analyse homework for the first hour and a half. Then, we work on a practical task in the second hour and a half. Each week, students will have a contest or a challenge (like Kaggle.com) to train a model for a particular task.
Grading
Mikhail Romanov, PhD, is a deep learning researcher and engineer. His experience includes deep learning for production, scientific computing and research, accompanied by teaching mathematics and machine learning in general.
His academic experience includes teaching courses at MIPT, HSE, Harbour Space Universities and online platforms. As a researcher, he has conducted research at the Technical University of Denmark, Mail.ru, Samsung Research, Quantori, and Yandex. In his research, his main areas of interest are depth estimation, optical flow, optimisation of neural networks, multi-task learning, self-supervised learning, LLMs and diffusion models. He has published papers on tomography, deep learning, scientific computing, computer vision, generative AI, and diffusion models.
See full profileApply for this course
Intro to Deep Learning
by Mikhail Romanov
Total hours
45 Hours
Dates
May 19 - Jun 06, 2025
Fee for single course
€1500
Fee for degree students
€750
How to secure your spot
Complete the form below to kickstart your application
Schedule your Harbour.Space interview
If successful, get ready to join us on campus
FAQ
Will I receive a certificate after completion?
Yes. Upon completion of the course, you will receive a certificate signed by the director of the program your course belonged to.
Do I need a visa?
This depends on your case. Please check with the Spanish or Thai consulate in your country of residence about visa requirements. We will do our part to provide you with the necessary documents, such as the Certificate of Enrollment.
Can I get a discount?
Yes. The easiest way to enroll in a course at a discounted price is to register for multiple courses. Registering for multiple courses will reduce the cost per individual course. Please ask the Admissions Office for more information about the other kinds of discounts we offer and what you can do to receive one.