Exploratory Analysis Exercises

# Solution Ex. 1.1
import re

def extract_year(moviename):
    m = re.search(r'\((\d*)\)', moviename)
    
    if m == None:
        return None
    
    return int(m.group(1))

# First step: extract release dates years from column moviename
df_mvl['release_year'] = df_mvl['moviename'].apply(extract_year)

df_mvl.hist(column='release_year')
df_mvl['release_year'].describe()

# Solution Ex. 1.2

df_mvl.groupby('userid')['rating'].mean().hist()

incremental_als.py

# Solution Ex. 2.2

def incremental_als(B, W, k, lambd=0.1, iter_per_row=50):
    n, m = B.shape
    Psi = np.zeros((n, k))
    Phi = np.zeros((k, m))

    for i in range(k):
        Psi[:, i] = np.random.rand(n)
        Phi[i, :] = np.random.rand(m)
        
        for j in range(iter_per_row):
            # Alternate optimization
            Psi[:, i] = fmin_l_bfgs_b(lambda psi_i: loss(np.hstack((Psi[:, :i], psi_i.reshape((n, 1)), Psi[:, i+1:])), Phi, B, W, lambd), 
                                     Psi[:, i], 
                                     approx_grad=1)[0]

            Phi[i, :] = fmin_l_bfgs_b(lambda phi_i: loss(Psi, np.vstack((Phi[:i, :], phi_i.reshape((1, m)), Phi[i+1:, :])), B, W, lambd), 
                                     Phi[i, :], 
                                     approx_grad=1)[0]
    return Psi, Phi