25-min talk about Machine Learning and a little bit of Deep Learning. Starts with some basic definitions (Supervised and Unsupervised Learning). Then, neural networks basic functionality is explained, ending up in Deep Learning and Convolutional Neural Networks.
Machine Learning Meetup that happened in Porto Alegre, Brazil.
3. How could we teach
machines in a way that they
could learn from
experience?
4. “A computer program is said to learn
from experience E with respect to some
class of tasks T and performance
measure P, if its performance at tasks in
T, as measured by P, improves with
experience E.
Tom M. Mitchell
5. “A computer program is said to learn
from experience E with respect to some
class of tasks T and performance
measure P, if its performance at tasks in
T, as measured by P, improves with
experience E.
Tom M. Mitchell
6. “A computer program is said to learn
from experience E with respect to some
class of tasks T and performance
measure P, if its performance at tasks in
T, as measured by P, improves with
experience E.
Tom M. Mitchell
7. “A computer program is said to learn
from experience E with respect to some
class of tasks T and performance
measure P, if its performance at tasks in
T, as measured by P, improves with
experience E.
Tom M. Mitchell
8. Example
Task T: Classify an e-mail as spam or not spam.
Performance measure P: Percentage of e-mails correctly
classified.
Training experience E: E-mails manually labeled by
humans.
13. Smaller than
others, with a
ball on the top:
group 0!
Not conic, nothing
on the top: group 1!
Higher than others,
with cross on the top:
group 2!
Unsupervised Learning
36. What happened?
• Major breakthrough in 2006 in the way deep architectures
were handled
• Unsupervised learning of representations is used to pre-train each
layer
• Unsupervised training of each layer at a time, on top of the previous
ones. The representation learned at each level is the input to the
next one.
• Use supervised learning to fine-tune all the layers
http://www.slideshare.net/perone/deep-learning-convolutional-neural-
networks
37. Besides that...
• Better hardware
• GPUs enable ~9x faster training compared to CPUs.
• High-quality datasets;
• New activation functions;
• Regularization methods.
41. Convolution
• In convolutional layers, all units are organized in feature maps;
• Mathematical term discrete convolution, which describes the
filtering operation performed by the feature map.
42. Convolution
"Deep Learning in a Nutshell" (https://devblogs.nvidia.com/parallelforall/deep-learning-nutshell-core-
concepts/)
43. Pooling
• Usually used after convolution
layers;
• Reduces spatial size of the
representation
• Reducing the quantity of needed
parameters and computation;
• Controlling overfitting.
"Convolutional Neural Networks for Visual Recognition" (http://cs231n.github.io/convolutional-
networks/)
44. Example – Max Pooling
"Convolutional Neural Networks for Visual Recognition" (http://cs231n.github.io/convolutional-
networks/)
50. But what if we could learn feature extractors
instead?
Traditional
Deep
Learning
Hand-crafted
feature extractor
Trainable Classifier
Trainable feature
extractor
Trainable Classifier