Deep Learning Essentials

Deep Learning Essentials

1. 7 - Convolutional Networks

   • Handling
   • Layer
     • Activation
     • Padding
       • k = (d-m+2p)/s+1
         • Wird noch korrigiert. Falsche Formel
     • Stride
       • Employes a vertical and horizonal axis
       • The number of steps a kernel is moved over the input activation matrix is called stride s.
   • Pooling
     • reduce the dimensionlity
   • Convolution
   • AlexNET - 2012

2. 6 - Model Assessment

3. 5 - Model Training

4. 4 - Weight Adaptation

5. 3 - Multilayer Perceptron

6. 2 - Linear Models

   • Supervised Learning
     • Learning with labels
       • each pattern x has a label information y
       • pair (x_i, y_i)
         • training set
         • ground truth
       • pair (x'_i, y'_i)
         • predict set
       • If the label is discrete, e.g., {0, 1} or {muffin, chihuahua}, the learning problem is called classification
         • classificatoin
       • if hoices is explored ol detection. continuous First, (y ∈ R) it is called regression.
         • regression
   • Linear Regression
     • found in natural and technical processes
     • The basic linear model (1)
       • x € IR
       • weight factor
         • w € IR called slope
       • parameter
         • b € IR called inter
       • linear relationship
     • Least Squares
       • With First, the least squares formulation, the coefficients can be derived. (2)
       • means squared error (MSE)
     • Linear Regression Coefficients
       • Weight and intercept can be mathematically derived as (3)
         • x Strich mens x_1 ... x_n and the same for the label y Strich
     • Example Fit, Illustration of linear model that is fiied to the patterns minimizing the MSE
   • Nearest Neighbors
     • K-nearest neighbors (kNN) searches for labels based on nieghborhoods in data space.

7. 1 - Introduction

   • A.I.
     • Intelligence is
       • learn from observations
         • others experiences
         • own experiences
     • related to
       • Data Science
       • Big Data
   • Technologies
     • deep learning
       • TFlearn
       • Keras
       • Tensorflow
     • mashine learning
       • scikit-learn
     • share on github
     • devlopment of scripts
       • Jupyter
     • Research
       • archivx
     • Powerfull Hardware
       • AWS
       • NVIDIA GRPUs

8. 8 - Neuroevolution

   • Genetic Alogrithm
     • mimicking biological evolution
       • Crossover
       • Mutation
       • Selection
     • Optimization
     • examples

9. 9 - Auto-encoder

10. 10 - Generative Adversarial Networks

Imagens e arquivos do mapa