Day 1 wazz up ai
- 9. Introduction
● How-to-AI Series will be held on 5 days, each day during the event you will
be introduced to different topics about AI and Machine Learning.
● Understand the common terms and be able to build an AI/ML product.
- 10. Timeline
Day 1: WazzUp AI 05/12/2021
Day 2: Build Up your own Neural Network 11/12/2021
Day 3: Take Up Convolutional Neural Network 12/12/2021
Day 4: Level Up your model 02/01/2022
Day 5: Meet Up with experts 09/01/2022
- 11. GDSC AI Challenge
Training your first AI model
● A competition hosted on Kaggle platform to help attendees revise what
they have learned in the first 3 workshops and build their own Machine
Learning models.
● Timeline: 12/12/2021 - 26/12/2021
● Total prizes: 3 million VND
- 13. how-to-AI Series: Unlock Potential
Day 1: WazzUp AI
Nguyễn Luật Gia Khôi
@giakhoi.nguyenluat
Nguyễn Hoàng Trung
@hoangtrung.nguyen
- 14. Outline
1. Break the ice
2. Introduction to How-to-AI Series: Unlock Potential
3. Introduction to AI/ML
4. What can AI/ML do ?
5. Introduction to Deep Learning
- 16. What is AI ?
An artificial intelligence (AI) is basically the mechanism to incorporate human
intelligence into machines through a set of rules (algorithm).
- 19. ANI
Artificial Narrow Intelligence
● Artificial narrow intelligence (ANI or narrow AI) refers to a computer’s ability
to perform a single task extremely well, such as crawling a webpage or
playing chess.
● Siri, Face ID and the Google Assistant, to self-driving cars, etc.
● The only form of AI that we have been able to develop so far. If you can
think about any form of AI that you know exists today, it is ANI.
- 23. “A year spent in artificial
intelligence is enough to make
one believe in God.”
- Alan Perlis -
- 24. What is Machine Learning ?
A subset of AI that uses statistical
learning algorithms to build smart
systems that can automatically learn
and improve without explicitly being
programmed.
- 25. Data
Data can be texts or numbers written
on papers, or it can be bytes and bits
inside the memory of electronic
devices, or it could be facts that are
stored inside a person’s mind.
- 26. Data in AI/ML
A good dataset? Some problems when choosing dataset ?
A good data set is one that has
either well-labeled fields and
members or a data dictionary
so you can re-label the data
yourself.
- 27. Data in AI/ML
A good dataset? Some problems when choosing dataset ?
● Wrong Data
● Missing Data
● Outliers in Data
● Redundancy in Data
● Unbalanced Data
● Lack of Variability in Data
- 39. Supervised vs Unsupervised
● Supervised learning: Learning on
labeled data. The goal is to learn a
general rule that maps inputs to
outputs
● Unsupervised learning: Learning
on unlabeled data. The goal is to
cluster the similar examples into
classes.
- 42. How can it know which step led to the failure?
Sparse
reward
Reward shaping
- 43. Application
● Supervised learning: regression, classification, etc.
● Unsupervised learning: segmentation, recommendation system, etc.
● Reinforcement learning: self-driving car, gaming, etc.
- 53. Reasons
● Training data is not cleaned and
contains noise.
● Complex model but small data set.
● Lack of features.
● The model is too simple
● Small size of dataset
For Overfitting: For Underfitting:
- 54. Solution
● Get/Generate more data
● Regularization
For Overfitting:
● Spend more time for preprocessing
data
● Increase the training duration
(# epochs)
● Increase model complexity
For Underfitting:
● Spend more time for feature
engineering (ML)
- 57. Why not see it ourselves?
teachablemachine.withgoogle.com