Estado del Arte de la IA
•Download as PPTX, PDF•
1 like•489 views
En esta sesión se hará un breve repaso histórico a los desarrollos en el campo de la inteligencia artificial, para después poder pasar a discutir el estado actual del campo, los nuevos desarrollos tanto en el entorno académico como el empresarial, y las líneas de investigación abiertas más interesantes.Repasaremos los papers más interesantes publicados este año en los campos de imagen artificial, procesado de texto y audio y aprendizaje por refuerzo, hablaremos del paso de las redes neuronales convolucionales a capsNets, y de como se están empleando estos nuevos desarrollos en clientes y empresas específicas.
More Related Content
Similar to Estado del Arte de la IA
Similar to Estado del Arte de la IA (20)
More from Plain Concepts
More from Plain Concepts (20)
Recently uploaded
Recently uploaded (20)
Estado del Arte de la IA
- 1. Un vistazo al Pasado y al Futuro de la IA
- 3. @PabloDoval palvarez@plainconcepts.com I work with code and data, but don't tell my mom; she thinks I'm a piano player in a whorehouse. Pablo Doval DATA PONTIFEX
- 4. “I’ll create a GUI interface using Visual Basic to see if I can track an IP address”
- 7. 7
- 8. PERCEPTRON
- 9. PERCEPTRON
- 14. EXPERT SYSTEMS?
- 16. MACHINE LEARNING
- 17. BACKPROPAGATIO N
- 18. DEEP LEARNING A Machine Learning technique
- 19. Image •Image Classification •Object Detection •Synthetical Generation of Images Sound •Fraud Detection •Defect Enhancement •Style Transfer Text •Information Retrieval •Knowledge Extraction •Sentiment Analysis •Style Transefer Signal Analysis •Time Series Analysis
- 24. 24 SHALL WE PLAY A LITTLE GAME?
- 25. A SIMPLE EXAMPLE 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 f(x) f(x) “X” “O”
- 26. 9 pixels 9 pixels 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 …… 81 values ( 𝑥) Ʃ Ʃ Ʃ Sum (weights x pixels) = 𝑤 𝑥 ⋅ 𝑥 Ʃ Sum (weights x pixels) =𝑤 𝑂 ⋅ 𝑥 Bias (𝑏) A SIMPLE EXAMPLE
- 27. NOW THAT YOU MENTION IT…
- 28. NOW THAT YOU MENTION IT… 𝑤
- 29. NOW THAT YOU MENTION IT… 𝑤 ∝ = 𝑤
- 30. NOW THAT YOU MENTION IT… 𝑤 ∝ = 𝑤
- 31. NOW THAT YOU MENTION IT… 𝑏 𝑏
- 32. NOW THAT YOU MENTION IT… 𝑏 𝑏
- 33. 9 pixels 9 pixels 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 …… 81 values ( 𝑥) Ʃ Ʃ 823,731.01 14,78 SIGMOID
- 34. 9 pixels 9 pixels 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 …… 81 values ( 𝑥) Ʃ Ʃ 0.78 0.32 SIGMOID
- 35. 9 pixels 9 pixels 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 …… 81 values ( 𝑥) Ʃ Ʃ 0.75 0.2 5 SOFTMAX
- 36. 9 pixels 9 pixels 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0.75 0.2 5 1 0 LOSS FUNCTION
- 37. 9 pixels 9 pixels 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0.75 0.2 5 1 0 LOSS FUNCTION
- 38. 9 pixels 9 pixels 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0.75 0.2 5 1 0 LOSS FUNCTION
- 39. GRADIENT DESCENT
- 40. GRADIENT DESCENT
- 41. REMEMBER THIS?
- 44. WHAT HAPPENED IN 2012? t2012 DNNs GPUs Deep Learning
- 46. 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 1 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 FILTERING
- 47. -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 1 -1 1 -1 1 -1 -1 1 -1 1 -1 1 -1 1 -1 1 1 1 1 1 1 1 1 1 1 -1 1 -1 1 1 1 -1 1 -1 1 1 -1 1 1 1 -1 1 1 Filter 1 Filter 2 Filter 3 1. 0 0.1 1 0.5 5 FILTERING …looking for common characteristics
- 48. 1 -1 -1 -1 1 -1 -1 -1 1 ⊗ -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 = 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 CONVOLUTION
- 49. 1 -1 -1 -1 1 -1 -1 -1 1 ⊗ -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 = 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 ⊗ = 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -1 -1 1 -1 1 -1 1 -1 -1 CONVOLUTION
- 50. 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 -1 -1 -1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 ⊗ = 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 1 -1 1 -1 1 -1 1 -1 1 0.33 -0.55 -0.11 -0.11 0.11 -0.55 0.33 -0.55 0.55 -0.55 0.33 -0.55 0.55 -0.55 0.11 -0.55 0.55 -0.77 0.55 -0.55 0.11 -0.11 0.33 -0.77 1.00 -0.77 0.33 -0.11 0.11 -0.55 0.55 -0.77 0.55 -0.55 0.11 -0.55 0.55 -0.55 0.33 -0.55 0.55 -0.55 0.33 -0.55 0.11 -0.11 0.11 -0.55 0.33 -1 -1 1 -1 1 -1 1 -1 -1
- 51. 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 POOLING
- 52. 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 POOLING
- 53. 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 1.00 1.00 0.33 0.55 0.55 0.33 1.00 1.00 1.00 0.33 0.11 0.55 0.33 1.00 1.00 0.55 0.33 0.55 0.55 0.33 0.55 1.00 1.00 0.33 0.55 0.11 0.33 1.00 1.00 1.00 0.33 0.55 0.55 0.33 1.00 1.00 MAX POOLING – STRIDE 1
- 54. 0.77 -0.11 0.11 0.33 0.55 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.11 -0.11 0.11 -0.11 1.00 -0.33 0.11 -0.11 0.55 0.33 0.33 -0.33 0.55 -0.33 0.33 0.33 0.55 -0.11 0.11 -0.33 1.00 -0.11 0.11 -0.11 0.11 -0.11 0.33 -0.11 1.00 -0.11 0.33 -0.11 0.55 0.33 0.11 -0.11 0.77 1.00 0.33 0.55 0.33 0.33 1.00 0.33 0.55 0.55 0.33 1.00 0.11 0.33 0.55 0.11 0.77 AVERAGE POOLING – STRIDE 2
- 55. WHY POOLING?
- 57. 1 -1 -1 -1 1 -1 -1 -1 1 -1 -1 1 -1 1 -1 1 -1 -1 1 -1 1 -1 1 -1 1 -1 1 Filter 1 Filter 2 Filter 3 WHAT IS MISSING?
- 58. AlexNet
- 60. 60 BASIC AUTOENCODER 𝑥 = ℎ 𝑊,𝑏 ≈ 𝑥
- 61. RNN & LONG SHORT TERM MEMORY
- 62. Evaluate the new opportunities offered to the breast cancer diagnosis by the latest advances in Deep Learning (deep neural models), extracting location, BIRADS classification and degree of confidence of each abnormality found in the mammography supplied as an input. Automated Mammogram BIRADS ClassifierAUTOENCODER, FASTRCNN
- 63. 63
- 64. A new method named Contextual Pyramid CNN (CP-CNN) is proposed here to generate density maps and influx estimations, by explicitely incorporating global and local context information. Composed of four modules: Global Context Estimator (GCE), Local Context Estimator (LCE), Density Map Estimator (DME) and a Fusion-CNN (F-CNN) convolutional network. Vishwanath A. Sindagi, Vishal M. Patel; The IEEE International Conference on Computer Vision (ICCV), 2017, pp. 1861-1870 Counting people in a crowd CONTEXTUAL PYRAMID CNN (CP-CNN)
- 68. By using a perceptual loss functions based on high- level features extracted from pretrained networks, networks for image transformation tasks can be trained, and by fine tuning the loss function different features can be kept for the source image and the style image. Justin Johnson, Alexandre Alahi, Li Fei-Fei; Perceptual Losses for Real-Time Style Transfer and Super- Resolution, 2016 Transferring style across images CONVOLUTIONAL NEURAL NETWORKS
- 69. HOW DOES THIS WORK?
- 71. By taking advantage of Generational Adversarial Networks, synthetic images based on the training data can be generated. Including an external array of features, the generated images can be tailored to a specific set of requirements. Jaime Deverall, Jiwoo Lee, Miguel Ayala; Using Generative Adversarial Networks to Design Shoes Using GANs to drive design decisions GENERATIVE ADVERSARIAL NETWORKS
- 74. HOW DOES THIS WORK?
- 75. 75
- 78. By using Generative Adversarial Networks, we are going to be able to upscale a pixelated image, and help the security enforcement team of our favourite TV show find the actual face of the criminal! Ledig, Theis, et al.; Photo-RealisticSingleImageSuper- ResolutionUsingaGenerativeAdversarial Network, 2017 Helping solve gruesome crimes SR-GAN AND SRRESNET, PIXELCNN
- 81. HOW DOES THIS WORK?
- 83. REMEMBER THIS?
- 91. MOVING FAST!
- 92. MOVING FAST!
- 95. Thanks and … See you soon! Thanks also to the organization Without whom this would not have been posible.
Editor's Notes
- Paul Stein and Nick Metropolis, playing chess in the MANIAC 1, in Los Alamos. First time a human was defeated in an “intellect game”. Rule based. They dressed way better than computer scientists today :)
- Frank Rossemblatt, 1956 400 photovoltaic cells as input Weights and bias as potentiometers, electrically rotated.
- Frank Rossemblatt, 1956 400 photovoltaic cells as input Wights and bias as potentiometers, electrically rotated.
- Marvin Minsky, Claude Shannon, Ray Solomonoff and other scientists 1956, Dartmouth Summer Research Project on Artificial Intelligence. Still dressing better than today :) Those were happy days…
- Ivaknenko developed MLP in 1965, but…
- Marvin Minksy and Seymour Papert.
- Sir Francis Galton and the linear regression.
- Geoff Hinton.