flowchart LR
subgraph Input["COUCHE D'ENTREE"]
x1((x1))
x2((x2))
end
subgraph Hidden["COUCHE CACHEE"]
h1((h1))
h2((h2))
h3((h3))
h4((h4))
end
subgraph Output["COUCHE DE SORTIE"]
y((y))
end
x1 --> h1 & h2 & h3 & h4
x2 --> h1 & h2 & h3 & h4
h1 & h2 & h3 & h4 --> y
y --> P[Prediction]
flowchart LR
subgraph Entrees
x1[x1]
x2[x2]
x3[x3]
end
subgraph Poids
w1[w1]
w2[w2]
w3[w3]
b[biais b]
end
x1 --- w1
x2 --- w2
x3 --- w3
w1 & w2 & w3 & b --> S["z = Σ wi·xi + b"]
S --> F["f(z)"]
F --> A[a = sortie]
Entrees: $\textcolor{#3498db}{x_1 = 0.5}$, $\textcolor{#e67e22}{x_2 = 0.8}$, $\textcolor{#27ae60}{x_3 = 0.2}$
Poids: $\textcolor{#3498db}{w_1 = 0.4}$, $\textcolor{#e67e22}{w_2 = 0.3}$, $\textcolor{#27ae60}{w_3 = 0.9}$, $\textcolor{#9B7AC4}{b = 0.1}$
$z = \textcolor{#3498db}{w_1 \cdot x_1} + \textcolor{#e67e22}{w_2 \cdot x_2} + \textcolor{#27ae60}{w_3 \cdot x_3} + \textcolor{#9B7AC4}{b}$
$z = \textcolor{#3498db}{0.4 \times 0.5} + \textcolor{#e67e22}{0.3 \times 0.8} + \textcolor{#27ae60}{0.9 \times 0.2} + \textcolor{#9B7AC4}{0.1}$
$z = \textcolor{#3498db}{0.20} + \textcolor{#e67e22}{0.24} + \textcolor{#27ae60}{0.18} + \textcolor{#9B7AC4}{0.10} = \textcolor{#F7E64D}{\mathbf{0.72}}$
$a = \sigma(\textcolor{#F7E64D}{0.72}) = \frac{1}{1 + e^{-\textcolor{#F7E64D}{0.72}}} = \textcolor{#e74c3c}{\mathbf{0.67}}$
Interpretation: Le neurone "s'active" a $\textcolor{#e74c3c}{67\%}$
- $\textcolor{#3498db}{Bleu}$: Entree 1 et son poids ($\textcolor{#3498db}{x_1, w_1}$)
- $\textcolor{#e67e22}{Orange}$: Entree 2 et son poids ($\textcolor{#e67e22}{x_2, w_2}$)
- $\textcolor{#27ae60}{Vert}$: Entree 3 et son poids ($\textcolor{#27ae60}{x_3, w_3}$)
- $\textcolor{#9B7AC4}{Violet}$: Biais ($\textcolor{#9B7AC4}{b}$)
- $\textcolor{#F7E64D}{Jaune}$: Somme ponderee z
- $\textcolor{#e74c3c}{Rouge}$: Activation finale a
flowchart LR
A["Donnees X"] --> B["Couche 1
z1=X·W1+b1
a1=relu(z1)"]
B --> C["Couche 2
z2=a1·W2+b2
y=sigmoid(z2)"]
C --> D["Prediction y"]
style A fill:#E5D7F5,color:#1A1A1A
style D fill:#FFF9D9,color:#1A1A1A
flowchart RL
E["Erreur
(y - ŷ)"] --> G2["Gradient W2
dL/dW2"]
G2 --> G1["Gradient W1
dL/dW1"]
G1 --> U["Mise a jour
W, b"]
style E fill:#C09CF0,color:#1A1A1A
style U fill:#F7E64D,color:#1A1A1A
flowchart TD
F["1. FORWARD
X → Couches → y_pred"]
F --> L["2. LOSS
Erreur = diff(y_pred, y_vrai)"]
L --> B["3. BACKWARD
dL/dW = contribution a l'erreur"]
B --> U["4. UPDATE
W = W - lr × gradient"]
U --> R{"Convergence?"}
R -->|Non| F
R -->|Oui| D["Modele entraine!"]
style F fill:#E5D7F5,color:#1A1A1A
style L fill:#FFF9D9,color:#1A1A1A
style B fill:#C09CF0,color:#1A1A1A
style U fill:#F7E64D,color:#1A1A1A
style D fill:#F7E64D,color:#1A1A1A