Merge pull request #15 from ankitaharwal/master

Fibonachi Series

Author: sanjaykr5

assignment_2/ankit/README.md

@@ -0,0 +1,27 @@
+## Problem Statement: 
+Train a model to print the first `n` elements of the Fibonacci sequence starting at integers `a` and `b`. Note that `a` < `b`, and `n` <= 100 are user supplied arguments. You may not use the standard summing operation to print the next element.Example I/O:
+Input: a=5, b=7, n=5
+Output: 5, 7, 12, 19, 31
+Input: a=7, b=41, n=6
+Output: 7, 41, 49, 90, 139, 229
+
+## Submission Guidelines
+1. Make a proper readme.md including the approach, dataset, model, experiments, results and steps to run the project.
+2. Include requirement.txt file.
+3. Make a predict.py file in which a user will input a,b,n and get the results.
+
+## Experiments
+1. I inserted little GUI in it because I recently learned little GUI and want to implement it
+
+## Approach
+1. We just have to make a model that can add two number.
+2. Our model will take a,b and return a+b
+3. We will make a function that will print fibonacci series
+ 
+## Dataset
+1. X has two features a,b and very big length
+2. y has a+b and same length as of X
+
+## Model
+1. We trained our model using simple linear regression using pytorch
+2. After training our model gives w=[[1],[1]] it is used for adding a,b

assignment_2/ankit/predict.py

@@ -0,0 +1,102 @@
+from torch.autograd import Variable
+import torch
+import tkinter as tk
+from tkinter import ttk
+import numpy as np
+import matplotlib.pyplot as plt
+import seaborn as sns
+import pandas as pd
+from sklearn.model_selection import train_test_split
+sns.set()
+
+win = tk.Tk()
+win.resizable(width=0, height=0)
+win.title("Fibonnachi")
+n_label = ttk.Label(win, text="Enter the n:")
+n_label.grid(row=0, column=0, sticky=tk.W)
+n_value = tk.IntVar()
+n_value.set(10)
+n_entrybox = ttk.Entry(win, width=14, textvariable=n_value)
+n_entrybox.grid(row=0, column=1)
+a_label = ttk.Label(win, text="a:")
+a_label.grid(row=1, column=0, sticky=tk.W)
+a_value = tk.IntVar()
+a_value.set(1)
+a_entrybox = ttk.Entry(win, width=14, textvariable=a_value)
+a_entrybox.grid(row=1, column=1)
+b_label = ttk.Label(win, text="b:")
+b_label.grid(row=2, column=0, sticky=tk.W)
+b_value = tk.IntVar()
+b_value.set(2)
+b_entrybox = ttk.Entry(
+    win, width=14, textvariable=b_value)
+b_entrybox.grid(row=2, column=1)
+n_out = 1
+n = 0
+A, B = [], []
+for i in range(1, 10):
+    for j in range(i, 10):
+        n += 1
+        A.append(i)
+        B.append(j)
+A, B = np.array(A), np.array(B)
+for i in range(n):
+    if A[i] > B[i]:
+        A[i], B[i] = B[i], A[i]
+data = pd.DataFrame({})
+data["A"] = A
+data["B"] = B
+X = np.array(data)
+y = np.array(data["A"]+data["B"]).reshape(-1, 1)
+X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=42)
+print("X", X[:5], "\n\n", "y", y[:5])
+dtype = torch.FloatTensor
+N, Din, Dout = n, 2, n_out
+x = Variable(torch.Tensor(X_train).type(dtype), requires_grad=False)
+x_test = Variable(torch.Tensor(X_test).type(dtype), requires_grad=False)
+y = Variable(torch.Tensor(y_train).type(dtype), requires_grad=False)
+y_test = Variable(torch.Tensor(y_test).type(dtype), requires_grad=False)
+w = Variable(torch.randn(Din, Dout).type(dtype), requires_grad=True)
+learning_rate = 0.00001
+train_loss_list = []
+test_loss_list = []
+for t in range(1000):
+    y_pred = x.mm(w)
+    y_pred_test = x_test.mm(w)
+    loss = (y_pred - y).pow(2).sum()
+    loss_test = (y_pred_test-y_test).pow(2).sum()
+    train_loss_list.append(loss.data)
+    test_loss_list.append(loss_test)
+    loss.backward()
+    w.data -= learning_rate * w.grad.data
+    w.grad.data.zero_()
+plt.plot(train_loss_list, label="train loss")
+plt.plot(test_loss_list, label="test_loss")
+plt.xlabel("number of iterations")
+plt.ylabel("cost")
+plt.legend()
+plt.show()
+
+
+def sum_new(a, b, w):
+    return round(float(torch.Tensor([[a, b]]).mm(w)[0][0].data))
+
+
+def fibo(w, a, b, n):
+    if n == 1:
+        return a
+    a_sum_b = sum_new(a, b, w)
+    return fibo(w, b, a_sum_b, n-1)
+
+
+def main():
+    a, b, n = a_value.get(), b_value.get(), n_value.get()
+    for i in range(1, n+1):
+        print(fibo(w, a, b, i), end=" ")
+    print("\n")
+
+
+submit_button = tk.Button(
+    win, text="Generate", command=main)
+submit_button.grid(row=3, columnspan=3)
+win.mainloop()

assignment_2/ankit/requirement

@@ -0,0 +1,6 @@
+1. python
+2. torch
+3. tkinter
+4. numpy
+5. matplotlib
+6. seaborn