1 of 19

Introduction to Deep Learning

Ph.D Tulio Campos

M.Sc. Daniel Farfán

2 of 19

What is Artificial Intelligence?

2

Artificial Intelligence

A vast array of techniques and methodologies aimed at enabling machines to perform tasks that typically require human intelligence

Artificial Intelligence

3 of 19

What is Machine Learning?

3

A subset of AI dedicated to developing algorithms that allow computers to learn from data and improve over time without explicit programming

Artificial Intelligence

Machine Learning

4 of 19

What is Deep Learning?

4

Is a specialized field within Machine Learning that focuses on algorithms inspired by the structure and function of the human brain's neural networks

Artificial Intelligence

Machine Learning

Deep Learning

5 of 19

Machine Learning vs Deep Learning

5

Feature	Machine Learning (ML)	Deep Learning (DL)
🧠 Relation to AI	A subset of Artificial Intelligence	A subset of Machine Learning
🔍 Feature Engineering	Requires manual feature engineering	Automatically learns relevant features
🧪 Amount of Data Needed	Can train on small to medium datasets	Requires large volumes of data
🧑‍🔧 Human Intervention	Needs human adjustments and corrections	Learns from environment and past mistakes
⏱️ Training Time	Shorter	Longer
🔁 Type of Relationships	Captures simple or linear patterns	Captures nonlinear, complex patterns
💰 Computational Cost	Low (efficient on CPU)	High (GPU recommended; CPU possible but inefficient)

6 of 19

Deep Learning Frameworks

6

Feature	PyTorch	Keras	TensorFlow
🎓 Learning Curve	Moderate	Low (easier)	High
🎨 Interface	Pythonic, closer to traditional programming	High-level, simple and user-friendly API	Complex, but simplified in TF 2.0
⚙️ Flexibility	High – allows detailed customization	Medium – simplifies common structures	High – but requires more code
🐞 Debugging	Simple, similar to standard Python	Limited by abstractions	Can be complex
🚀 Advantages	• Fine control of the process • Better for experimental research • More intuitive programming	• Rapid prototyping • Less code to get started • Ideal for beginners	• Complete ecosystem • Better performance in production • Tools for large datasets
❗ Limitations	• More code for simple tasks • Fewer built-in production tools	• Less flexible for new/complex models • Depends on TensorFlow	• Greater initial complexity • Less intuitive API
🎯 Ideal for Users Who…	Have some programming experience and need to customize models for research	Are new to DL and want quick results with minimal code	Work with large datasets and need scalable, end-to-end solutions

7 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

7

Types of Learning

8 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

8

Types of Learning

Supervised Learning

9 of 19

Supervised Learning

Uses data sets tagged to train algorithms to predict results and recognize patterns

Key characteristics:

It receives input-output pairs.
The goal is to predict outputs for new inputs.

9

Labeled Data

Circle

Triangle

Square

Labels

Model Training

10 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

10

Types of Learning

Supervised Learning

11 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

11

Types of Learning

Supervised Learning

Semi-supervised Learning

12 of 19

Semi-supervised Learning

The model learns from a small amount of labeled data combined with a large amount of unlabeled data

Key characteristics:

Uses both labeled and unlabeled data.
Bridges the gap between supervised and unsupervised learning.

12

Large amount of unlabeled data

Circle

Triangle

Square

Small amount of labeled data

Model Training

13 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

13

Types of Learning

Supervised Learning

Semi-supervised Learning

14 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

14

Types of Learning

Supervised Learning

Semi-supervised Learning

Unsupervised Learning

15 of 19

Unsupervised Learning

Uses unlabeled data and can discover patterns and statistics without any explicit guidance or instruction

Key characteristics:

Only inputs are provided (no labels).
The goal is to group, reduce dimensions, or detect structure.

15

Raw Data

Model Training

16 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

16

Types of Learning

Supervised Learning

Unsupervised Learning

Semi-supervised Learning

17 of 19

Types of Learning in Artificial Intelligence

AI learns from data.

The type of learning depends on:

The kind of data available.
The goal to be achieved.

17

Types of Learning

Supervised Learning

Unsupervised Learning

Semi-supervised Learning

Reinforcement Learning

18 of 19

Reinforcement Learning

The model learns by trial and

error to maximize a reward.

Key characteristics:

An agent makes decisions in an environment.
It receives rewards or penalties based on behavior.

18

Input Raw Data

Environment

Agent

Reward

Action

State

Selection of Algorithm

Output

19 of 19

Types of Learning in Artificial Intelligence

19

Type	Requires Labels	Main Goal	Example in Biology
Supervised	✅ Yes	Prediction	Image-based diagnosis
Unsupervised	❌ No	Pattern discovery	Gene profile clustering
Semi-Supervised	📄 Few	Improve prediction + discovery	Classification with limited data
Reinforcement	⚠️ Not direct	Optimal decision-making	Biological Sequence Design