16. Pre Processing the Data
Use Text classification:
- Support Vector Machine (SVM)
- Convolution Neural Network (CNN)
In Order to FILTER the useful data.
Build time-series of targeted products.
18. 2. Prediction H Bakir, G Chniti, H Zaher
E-Commerce price forecasting using LSTM neural networks
International Journal of Machine Learning and Computing 8 (2), 169-174
30. Inspiration of Convolution Networks
“Receptive fields, binocular interaction and functional architecture in the cat's visual cortex” 1968 - D. H. Hubel and T.
N. Wiesel
30
31. New Idea: Known unknowns => unknown unknowns
31
Neocognitron - Fukushima 1980
57. Tensorflow
1. Powerful Deep Learning Framework
2. Keras
3. The approach is optional (Tensorflow Eager)
4. Tensorboard
57
58. CNN in Keras
58
Convolution
Max Pooling
Convolution
Max Pooling
Input
1 x 28 x 28
25 x 26 x 26
25 x 13 x 13
50 x 11 x 11
50 x 5 x 5
How many
parameters for each
filter?
How many
parameters
for each filter?
9
225=
25x9
59. CNN in Keras
59
Convolution
Max Pooling
Convolution
Input
1 x 28 x 28
25 x 26 x 26
25 x 13 x 13
50 x 11 x 11
Flattened
1250
Fully connected
feedforward network
Output
50 x 5 x 5
Max Pooling
60. CNN LOSS FUNCTIONS
Can be categorized in to the following categories:
1. Binary Classification(SVM hinge loss, Squared hinge
loss).
2. Identity Verification(Contrastive loss).
3. Multi-class Classification (Softmax loss, Expectation loss).
4. Regression (SSIM,`1 error, Euclidean loss)
60