BatchLinearRegor

A Py3 code that implements batch linear regressor using gradient descent.

Math path

Write a Python code that implements batch linear regressor using gradient descent.

Parameters:

• threshold - The threshold, that the change in error has to fall below, before the algorithm terminates.
• data - The location of the data file (e.g. /media/data/yacht.csv).
• eta - The learning rate of the gradient descent approach.

Start your program in the following way: python3 student.py --data random.csv --eta 0.0001 --threshold 0.0001

Output

Bonus

def regressor(dataset,eta,t):
    x, y, w = mat(dataset)
    costmat,res=[],[]
    key=0
    while True:
        currres=[]
        currres.append(key)
        key += 1
        cost = sqrerror(x,y,w)
        for ol in w:
            for il in ol:
                currres.append(il)
        currres.append(cost)
        costmat.append(cost)
        res.append(currres)
        w = w-(eta)*np.sum((x@w.T-y)*x,axis=0) #Gradient Calculation 

        if key>1:
            if (costmat[-2]-costmat[-1]) <= t:
                break
    return res

Standard release

A build of the same problem statement with only using standard libraries is available at the release section.

Run Dockerfile

docker run -it python-regor1 python3 bregressorv2.py --data random1.csv --eta 0.0001 --threshold 0.0001