add perceptron

.gitignore

@@ -5,4 +5,6 @@ __pycache__
 
 **/.DS_Store
 
-notebooks/
+notebooks/
+
+.idea/

models/Perceptron.py

@@ -0,0 +1,34 @@
+import streamlit as st
+from sklearn.linear_model import Perceptron
+
+
+def lp_param_selector():
+ eta0 = st.slider("learning_rate", 0.001, 10.0, step=0.001, value=1.0)
+ max_iter = st.number_input("max_iter", 100, 2000, step=50, value=100)
+
+ penalty = st.selectbox("penalty", options=["None", "l2", "l1", "elasticnet"])
+
+ if penalty in ["l2", "l1", "elasticnet"]:
+ alpha = st.slider("alpha", 0.00001, 0.001, step=0.00001, value=0.0001)
+ else:
+ alpha = 0.0001
+
+ early_stopping = st.checkbox('early_stopping', value=False)
+
+ if early_stopping:
+ validation_fraction = st.number_input("validation_fraction", 0.0, 1.0, step=0.05, value=0.1)
+ n_iter_no_change = st.number_input("n_iter_no_change", 2, 100, step=1, value=5)
+ else:
+ validation_fraction = 0.1
+ n_iter_no_change = 5
+
+ params = {"eta0": eta0, 
+ "max_iter": max_iter, 
+ "penalty": penalty, 
+ "alpha": alpha, 
+ "early_stopping": early_stopping,
+ "validation_fraction": validation_fraction, 
+ "n_iter_no_change": n_iter_no_change}
+
+ model = Perceptron(**params)
+ return model

models/utils.py

@@ -7,6 +7,7 @@
  "K Nearest Neighbors": "from sklearn.neighbors import KNeighborsClassifier",
  "Gaussian Naive Bayes": "from sklearn.naive_bayes import GaussianNB",
  "SVC": "from sklearn.svm import SVC",
+ "Perceptron": "from sklearn.linear_model import Perceptron",
 }
 
 
@@ -19,6 +20,7 @@
  "K Nearest Neighbors": "https://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html",
  "Gaussian Naive Bayes": "https://scikit-learn.org/stable/modules/naive_bayes.html",
  "SVC": "https://scikit-learn.org/stable/modules/generated/sklearn.svm.SVC.html",
+ "Perceptron": "https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.Perceptron.html",
 }
 
 
@@ -63,4 +65,10 @@
  - They provide different type of kernel functions
  - They require careful normalization 
  """,
+ "Perceptron": """
+ - Perceptron is the simples form of Artificial Neural Network
+ - Single-layer Perceptrons require linearly separable data, otherwise it does not converge
+ - Multilayer Perceptrons (MLP) are feedforward neural networks with two or more layers which have the greater 
+ processing power and can process non-linearly separable data. 
+ """,
 }

utils/ui.py

@@ -10,6 +10,7 @@
 from models.KNearesNeighbors import knn_param_selector
 from models.SVC import svc_param_selector
 from models.GradientBoosting import gb_param_selector
+from models.Perceptron import lp_param_selector
 
 from models.utils import model_imports
 from utils.functions import img_to_bytes
@@ -83,6 +84,7 @@ def model_selector():
  "K Nearest Neighbors",
  "Gaussian Naive Bayes",
  "SVC",
+ "Perceptron",
  ),
  )
 
@@ -110,6 +112,9 @@ def model_selector():
  elif model_type == "Gradient Boosting":
  model = gb_param_selector()
 
+ elif model_type == "Perceptron":
+ model = lp_param_selector()
+
  return model_type, model