2023-12-22 16:39:03 +00:00
---
tags:
- ai
---
2023-06-01 08:11:37 +01:00
![deep-digit-classification ](../../img/deep-digit-classification.png )
2023-05-26 06:37:13 +01:00
2023-06-08 17:52:09 +01:00
OCR [Classification ](../Classification/Classification.md )
2023-05-26 06:37:13 +01:00
# Loss Function
Objective Function
2023-06-01 08:11:37 +01:00
- [Back-Propagation ](MLP/Back-Propagation.md )
2023-05-26 06:37:13 +01:00
- Difference between predicted and target outputs
2023-06-01 08:11:37 +01:00
![deep-loss-function ](../../img/deep-loss-function.png )
2023-05-26 06:37:13 +01:00
- Test accuracy worse than train accuracy = overfitting
2023-06-06 11:48:49 +01:00
- [Dense ](MLP/MLP.md ) = [fully connected ](MLP/MLP.md )
2023-05-26 06:37:13 +01:00
- Automates feature engineering
2023-06-01 08:11:37 +01:00
![ml-dl ](../../img/ml-dl.png )
2023-05-26 06:37:13 +01:00
These are the two essential characteristics of how deep learning learns from data: the incremental, layer-by-layer way in which increasingly complex representations are developed, and the fact that these intermediate incremental representations are learned jointly, each layer being updated to follow both the representational needs of the layer above and the needs of the layer below. Together, these two properties have made deep learning vastly more successful than previous approaches to machine learning.
# Steps
Structure defining
Compilation
- Loss function
- Metric of difference between output and target
- Optimiser
- How network will update
- Metrics to monitor
- Testing and training
Data preprocess
- Reshape input frame into linear array
- Categorically encode labels
Fit
Predict
Evaluate
# Data Structure
2023-06-01 08:11:37 +01:00
- [Tensor ](../../Maths/Tensor.md ) flow = channels last
2023-05-26 06:37:13 +01:00
- (samples, height, width, channels)
- Vector data
2023-06-01 08:11:37 +01:00
- 2D [tensors ](../../Maths/Tensor.md ) of shape (samples, features)
2023-05-26 06:37:13 +01:00
- Time series data or sequence data
2023-06-01 08:11:37 +01:00
- 3D [tensors ](../../Maths/Tensor.md ) of shape (samples, timesteps, features)
2023-05-26 06:37:13 +01:00
- Images
2023-06-01 08:11:37 +01:00
- 4D [tensors ](../../Maths/Tensor.md ) of shape (samples, height, width, channels) or (samples, channels, height, Width)
2023-05-26 06:37:13 +01:00
- Video
2023-06-01 08:11:37 +01:00
- 5D [tensors ](../../Maths/Tensor.md ) of shape (samples, frames, height, width, channels) or (samples, frames, channels , height, width)
2023-05-26 06:37:13 +01:00
2023-06-01 08:11:37 +01:00
![photo-tensor ](../../img/photo-tensor.png )
![matrix-dot-product ](../../img/matrix-dot-product.png )