Elstuhn| gradient-descent| Added main.py and readme (#246)

Add gradient descent implementation

Author: Elstuhn

gradient-descent/README.md

@@ -0,0 +1,2 @@
+# Made by [Elstuhn](https://github.com/Elstuhn)
+# Gradient Descent with Approximation in Python

gradient-descent/main.py

@@ -0,0 +1,53 @@
+### Made By Elston
+### https://github.com/Elstuhn
+### Gradient descent using approximation
+
+from functools import cache 
+import sympy
+
+@cache
+def findgrad(x : float, f, step : float):
+    return (f(x+step)-f(x))/step
+
+def cleanfunc(func : str):
+    x = sympy.Symbol("x")
+    try:
+        f = sympy.sympify(func)
+        return sympy.lambdify(x, f)
+    except:
+        try:
+            func = func.replace("^", "**")
+            f = sympy.sympify(func)
+            return sympy.lambdify(x, f)
+        except:
+            pass
+    inds = []
+    func = list(func)
+    for i, j in enumerate(func):
+        if all([
+            j == "x",
+            func[i-1] != "(",
+            func[i-1].isdigit(),
+            i != 0,
+        ]):
+            inds.append(i)
+    
+    prev = 0
+    result = ""
+    for index in inds:
+        result += "".join(func[prev:index])+"*"
+        prev = index
+    result += "".join(func[prev:])
+    f = sympy.sympify(result)
+    return sympy.lambdify(x, f)        
+
+
+def grad_desc(x : float, func : str, step : int = 0.05, lr : float = 3e-4, threshold : float = 0.001):
+    """
+    Trains x by approximating the gradient and checks it against the threshold
+    """
+    f = cleanfunc(func)
+    while findgrad(x, f, step) > threshold:
+        x -= findgrad(x, f, step)*lr
+        
+    return x