Learn computer vision fundamentals with the famous MNIST data
MNIST dataset:
- `test.csv` - https://www.kaggle.com/c/3004/download/test.csv
- `train.csv` - https://www.kaggle.com/c/3004/download/train.csv
The data files contain gray-scale images of hand-drawn digits, from zero through nine.
Each image is 28 pixels in height and 28 pixels in width, for a total of 784 pixels in total. Each pixel has a single pixel-value associated with it, indicating the lightness or darkness of that pixel, with higher numbers meaning darker. This pixel-value is an integer between 0 and 255, inclusive.
The training data set, (train.csv), has 785 columns. The first column, called "label", is the digit that was drawn by the user. The rest of the columns contain the pixel-values of the associated image.
The test data set, (test.csv), is the same as the training set, except that it does not contain the "label" column.
.
├── ANN.ipynb
├── train_test.py
├── ANN_class.py
└── README.md
# of layers: 2
# neurals for each layer: 64, 64
Epoch: 20
Alpha: 0.1
Training Error: 0.305111104410624
>epoch=0, lr=0.100, error=0.305, accuracy=0.9159
Training Error: 0.11553265572269808
Training Error: 0.08597945220389482
Training Error: 0.06923299499781688
Training Error: 0.05804369756110391
Training Error: 0.04971688542357365
>epoch=5, lr=0.100, error=0.050, accuracy=0.9727
Training Error: 0.04316807849710413
Training Error: 0.03792529986282915
Training Error: 0.033453521099773405
Training Error: 0.02957115827923241
Training Error: 0.026376685771993025
>epoch=10, lr=0.100, error=0.026, accuracy=0.9846
Training Error: 0.023722349178411532
Training Error: 0.02143442066478345
Training Error: 0.019500904765450598
Training Error: 0.01789154038737929
Training Error: 0.016455202489956695
>epoch=15, lr=0.100, error=0.016, accuracy=0.9854
Training Error: 0.015221036176883038
Training Error: 0.014259407447295421
Training Error: 0.013333599632849448
Training Error: 0.012566579491967657
CPU times: user 6min 13s, sys: 44.2 s, total: 6min 57s
Wall time: 4min 17s
Confusion matrix can be very helpfull to see your model drawbacks.
I plot the confusion matrix of the validation results.
