|
|
```python
|
|
|
#!pip3 install tensorflow
|
|
|
import tensorflow as tf
|
|
|
import numpy as np
|
|
|
import matplotlib.pyplot as plt
|
|
|
```
|
|
|
|
|
|
|
|
|
```python
|
|
|
# data for training the ann mode
|
|
|
# option 1:
|
|
|
celsius = np.array([-40, -10, 0, 8, 15, 22, 38], dtype=float)
|
|
|
fahrenheit = np.array([-40, 14, 32, 46, 59, 72, 100], dtype=float)
|
|
|
|
|
|
# option 2: (X°C x 9/5) + 32 = 41 °F
|
|
|
points = 100
|
|
|
np.random.seed(99)
|
|
|
dataIn = np.linspace (-40,60, points)
|
|
|
target = dataIn*9/5 + 32 +4*np.random.randn(points)
|
|
|
|
|
|
plt.plot(celsius, fahrenheit, 'or', label='data-set 1')
|
|
|
plt.plot(dataIn, target, '.b', alpha=0.3, label='data-set 2')
|
|
|
plt.legend()
|
|
|
plt.grid()
|
|
|
plt.show()
|
|
|
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
```python
|
|
|
from tensorflow.keras.models import Sequential # ANN type
|
|
|
from tensorflow.keras.layers import Dense, Input # All nodes connected
|
|
|
|
|
|
# NN definition
|
|
|
hn=2
|
|
|
model = Sequential()
|
|
|
model.add(Input(shape=(1,), name='input'))
|
|
|
model.add(Dense(hn, activation='linear', name='hidden'))
|
|
|
model.add(Dense(1, activation='linear', name='output'))
|
|
|
model.summary()
|
|
|
```
|
|
|
|
|
|
|
|
|
<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace"><span style="font-weight: bold">Model: "sequential_1"</span>
|
|
|
</pre>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace">┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
|
|
|
┃<span style="font-weight: bold"> Layer (type) </span>┃<span style="font-weight: bold"> Output Shape </span>┃<span style="font-weight: bold"> Param # </span>┃
|
|
|
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
|
|
|
│ hidden (<span style="color: #0087ff; text-decoration-color: #0087ff">Dense</span>) │ (<span style="color: #00d7ff; text-decoration-color: #00d7ff">None</span>, <span style="color: #00af00; text-decoration-color: #00af00">2</span>) │ <span style="color: #00af00; text-decoration-color: #00af00">4</span> │
|
|
|
├─────────────────────────────────┼────────────────────────┼───────────────┤
|
|
|
│ output (<span style="color: #0087ff; text-decoration-color: #0087ff">Dense</span>) │ (<span style="color: #00d7ff; text-decoration-color: #00d7ff">None</span>, <span style="color: #00af00; text-decoration-color: #00af00">1</span>) │ <span style="color: #00af00; text-decoration-color: #00af00">3</span> │
|
|
|
└─────────────────────────────────┴────────────────────────┴───────────────┘
|
|
|
</pre>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace"><span style="font-weight: bold"> Total params: </span><span style="color: #00af00; text-decoration-color: #00af00">7</span> (28.00 B)
|
|
|
</pre>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace"><span style="font-weight: bold"> Trainable params: </span><span style="color: #00af00; text-decoration-color: #00af00">7</span> (28.00 B)
|
|
|
</pre>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<pre style="white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace"><span style="font-weight: bold"> Non-trainable params: </span><span style="color: #00af00; text-decoration-color: #00af00">0</span> (0.00 B)
|
|
|
</pre>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
```python
|
|
|
### veri important note implement a python code
|
|
|
# to show the ANN model connection using ascii
|
|
|
```
|
|
|
|
|
|
|
|
|
```python
|
|
|
from tensorflow.keras.optimizers import Adam
|
|
|
|
|
|
#hyper parameters
|
|
|
epoch = 500
|
|
|
lr = 0.01
|
|
|
hn = 2 # hidden nodes
|
|
|
tf.random.set_seed(42) # For TensorFlow
|
|
|
|
|
|
|
|
|
model.compile(optimizer=Adam(lr), loss='mean_squared_error')
|
|
|
print("Starting training ...")
|
|
|
historial = model.fit(dataIn, target, epochs=epoch, verbose=False,)
|
|
|
print("Model trainned!")
|
|
|
```
|
|
|
|
|
|
Starting training ...
|
|
|
Model trainned!
|
|
|
|
|
|
|
|
|
|
|
|
```python
|
|
|
predict = model.predict(dataIn)
|
|
|
plt.plot(dataIn, predict, ':r', label='estimated')
|
|
|
plt.plot(dataIn,target, '.b', label='real', alpha=0.4)
|
|
|
plt.legend()
|
|
|
plt.grid()
|
|
|
plt.show()
|
|
|
```
|
|
|
|
|
|
[1m4/4[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 4ms/step
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
```python
|
|
|
# Get weights
|
|
|
for layer in model.layers:
|
|
|
weights = layer.get_weights()
|
|
|
print(f"Layer: {layer.name}")
|
|
|
print(f" Weights (Kernel): {weights[0].shape} \n{weights[0]}")
|
|
|
print(f" Biases: {weights[1].shape} \n{weights[1]}")
|
|
|
```
|
|
|
|
|
|
Layer: hidden
|
|
|
Weights (Kernel): (1, 2)
|
|
|
[[-0.27738443 0.7908125 ]]
|
|
|
Biases: (2,)
|
|
|
[-8.219968 6.714554]
|
|
|
Layer: output
|
|
|
Weights (Kernel): (2, 1)
|
|
|
[[-1.9934888]
|
|
|
[ 1.5958738]]
|
|
|
Biases: (1,)
|
|
|
[5.1361823]
|
|
|
|
|
|
|
|
|
# Testing the model
|
|
|
|
|
|
|
|
|
```python
|
|
|
inTest = np.array([100])
|
|
|
model.predict(inTest)
|
|
|
```
|
|
|
|
|
|
[1m1/1[0m [32m━━━━━━━━━━━━━━━━━━━━[0m[37m[0m [1m0s[0m 87ms/step
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
array([[213.73816]], dtype=float32)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
```python
|
|
|
# Do the Maths:
|
|
|
inTest = np.array(inTest)
|
|
|
whi = np.array([[-0.27738443, 0.7908125 ]])
|
|
|
bh = np.array([-8.219968, 6.714554])
|
|
|
Oh = np.dot(inTest,whi)+bh
|
|
|
who = np.array([[-1.9934888],[ 1.5958738]])
|
|
|
bo = np.array([5.1361823])
|
|
|
Oo = np.dot(Oh,who)+bo
|
|
|
Oo
|
|
|
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
array([213.73814765])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
```python
|
|
|
def generate_ascii_ann(model):
|
|
|
ascii_diagram = "\nArtificial Neural Network Architecture:\n"
|
|
|
|
|
|
for i, layer in enumerate(model.layers):
|
|
|
weights = layer.get_weights()
|
|
|
|
|
|
# Determine layer type and number of neurons
|
|
|
if isinstance(layer, Dense):
|
|
|
input_dim = weights[0].shape[0] # Number of inputs
|
|
|
output_dim = weights[0].shape[1] # Number of neurons
|
|
|
|
|
|
ascii_diagram += f"\nLayer {i+1}: {layer.name} ({layer.__class__.__name__})\n"
|
|
|
ascii_diagram += f" Inputs: {input_dim}, Neurons: {output_dim}\n"
|
|
|
ascii_diagram += f" Weights Shape: {weights[0].shape}\n"
|
|
|
|
|
|
if len(weights) > 1: # If bias exists
|
|
|
ascii_diagram += f" Biases Shape: {weights[1].shape}\n"
|
|
|
|
|
|
# ASCII representation of neurons
|
|
|
ascii_diagram += " " + " o " * output_dim + " <- Output Neurons\n"
|
|
|
ascii_diagram += " | " * output_dim + "\n"
|
|
|
ascii_diagram += " " + " | " * input_dim + " <- Inputs\n"
|
|
|
|
|
|
return ascii_diagram
|
|
|
|
|
|
# Generate and print the ASCII diagram
|
|
|
ascii_ann = generate_ascii_ann(model)
|
|
|
print(ascii_ann)
|
|
|
```
|
|
|
|
|
|
|
|
|
Artificial Neural Network Architecture:
|
|
|
|
|
|
Layer 1: hidden (Dense)
|
|
|
Inputs: 1, Neurons: 2
|
|
|
Weights Shape: (1, 2)
|
|
|
Biases Shape: (2,)
|
|
|
o o <- Output Neurons
|
|
|
| |
|
|
|
| <- Inputs
|
|
|
|
|
|
Layer 2: output (Dense)
|
|
|
Inputs: 2, Neurons: 1
|
|
|
Weights Shape: (2, 1)
|
|
|
Biases Shape: (1,)
|
|
|
o <- Output Neurons
|
|
|
|
|
|
|
| | <- Inputs
|
|
|
|
|
|
|
|
|
|
|
|
```mermaid
|
|
|
graph LR
|
|
|
I1((I_1)) --> H1((H_1)) & H2((H_1))
|
|
|
H1 & H2 --> O1((O_1)) & O2((O_2))
|
|
|
```
|