def compute_perplexities(vals, test_data, smoothing):
return [np.exp(-log_likelihood(test_data, wordprobs(test_data, val))
/ test_data.sum()) for val in vals]
def compute_perplexities(vals, test_data, smoothing):
return [
np.exp(-log_likelihood(test_data, wordprobs(test_data, val)) /
test_data.sum()) for val in vals
]
def compute_loglikes(vals, train_data, smoothing):
return [log_likelihood(train_data, wordprobs(train_data, val))
for val in vals]
def compute_loglikes(vals, train_data, smoothing):
return [
log_likelihood(train_data, wordprobs(train_data, val)) for val in vals
]