Deep Learning (DEEP-0001)�

Prof. André E. Lazzaretti

lazzaretti@utfpr.edu.br

https://sites.google.com/site/andrelazzaretti/graduate-courses/deep-learning-cpgei/2025

Week 1 - Introduction

Book

http://udlbook.com

​

Slides based on the material from the book and the CM20315 course (University of Bath)

Supervised learning

• Define a mapping from input to output
• Learn this mapping from paired input/output data examples

What is a supervised learning model?

• An equation relating input (age) to output (height)
• Search through family of possible equations to find one that fits training data well

What is a supervised learning model?

• Deep neural networks are just a very flexible family of equations
• Fitting deep neural networks = “Deep Learning”

Regression

• Univariate regression problem (one output, real value)
• Fully connected network

Graph regression

• Multivariate regression problem (>1 output, real value)
• Graph neural network

Text classification

• Binary classification problem (two discrete classes)
• Transformer network

Music genre classification

• Multiclass classification problem (discrete classes, >2 possible values)
• Recurrent neural network (RNN)

Image classification

• Multiclass classification problem (discrete classes, >2 possible classes)
• Convolutional network

Image segmentation

• Multivariate binary classification problem (many outputs, two discrete classes)
• Convolutional encoder-decoder network

Depth estimation

• Multivariate regression problem (many outputs, continuous)
• Convolutional encoder-decoder network

Pose estimation

• Multivariate regression problem (many outputs, continuous)
• Convolutional encoder-decoder network

Translation

Multiple valid translations from an English sentence to French.

Image captioning

Multiple images that are compatible with any caption.

Image generation from text

What do these last three examples have in common?

• Very complex relationship between input and output
• Sometimes, there may be many possible valid answers
• But outputs (and sometimes inputs) obey rules: can be learned without the need for output labels.

Language obeys grammatical rules

Natural images also have “rules”

Unsupervised Learning

• Learning about a dataset without labels
• Clustering
• Generating new examples
• Filling in missing data

DeepCluster: Deep Clustering for Unsupervised Learning of Visual Features (Caron et al., 2018)

Unsupervised Learning

• Learning about a dataset without labels
• e.g., clustering
• Generative models can create examples
• e.g., generative adversarial networks
• PGMs learn distribution over data
• e.g., variational autoencoders,
• e.g., normalizing flows,
• e.g., diffusion models

​

Generative models

Latent variables

Interpolation

Conditional synthesis

Reinforcement learning

• A set of states
• A set of actions
• A set of rewards

​

• Goal: take actions to change the state so that you receive rewards

​

• You don’t receive any data – you have to explore the environment yourself to gather data as you go

​

Example: chess

• States are valid states of the chess board
• Actions at a given time are valid possible moves
• Positive rewards for taking pieces, negative rewards for losing them

​

Why is this difficult?

• Stochastic
• Make the same move twice, the opponent might not do the same thing
• Rewards also stochastic (opponent does or doesn’t take your piece)
• Temporal credit assignment problem
• Did we get the reward because of this move? Or because we made good tactical decisions somewhere in the past?
• Exploration-exploitation trade-off
• If we found a good opening, should we use this?
• Or should we try other things, hoping for something better?

Exploration-exploitation trade-off

Landmarks in Deep Learning

• 1958 Perceptron (Simple `neural’ model)
• 1986 Backpropagation (Practical Deep Neural networks)
• 1989 Convolutional networks (Supervised learning)
• 2012 AlexNet Image classification (Supervised learning)
• 2014 Generative adversarial networks (Unsupervised learning)
• 2014 Deep Q-Learning -- Atari games (Reinforcement learning)
• 2016 AlphaGo (Reinforcement learning)
• 2017 Machine translation (Supervised learning)
• 2019 Language models ((Un)supervised learning)
• 2022 Dall-E2 Image synthesis from text prompts ((Un)supervised learning)
• 2022 ChatGPT ((Un)supervised learning)
• 2023 GPT4 Multimodal model ((Un)supervised learning)
• …

​

​

2018 Turing award winners

This course

Deep neural networks

How to train them

How to measure their performance

How to make that performance better

This course

Networks specialized in images

Image classification, object detection

Image segmentation

Pose estimation

This course

Networks specialized to text

Text generation

Automatic translation

ChatGPT

This course

Generative learning (unsupervised)

Generating random cats!

This course

Play Atari games, AlphaGo

Robots that learn how to walk