CODEWITHSUNDEEP
pythonstan
user5779223
user5779223

Reputation: 1490

Fail to implement the pmf model in pystan

I am a newbie to both stan and machine learning. Now I'd like to implement the pmf model. Here is part of my code:

pmf_cod="""
data {

int<lower=0> K; //number of factors
int<lower=0> N; //number of user
int<lower=0> M; //number of item
int<lower=0> D; //number of observation
int<lower=0> D_new; //number of pridictor 
int<lower=0, upper=N> ii[D]; //item 
int<lower=0, upper=M> jj[D]; //user
int<lower=0, upper=N> ii_new[D_new]; // item
int<lower=0, upper=N> jj_new[D_new]; // user
real<lower=0, upper=5> r[D]; //rating
real<lower=0, upper=5> r_new[D_new]; //pridict rating

}

parameters {
row_vector[K] i[M]; // item profile
row_vector[K] u[N]; // user profile
real<lower=0> alpha;
real<lower=0> alpha_i;
real<lower=0> alpha_u;

}

transformed parameters {
matrix[N,M] I; // indicator variable
I <- rep_matrix(0, N, M);
for (d in 1:D){
    I[ii[d]][jj[d]] <- 1;
}
}

model {
for (d in 1:D){
    r[d] ~ normal(sum(u[jj[d]]' * i[ii[d]]), 1/alpha);
}

for (n in 1: N){
    u[n] ~ normal(0,(1/alpha_u) * I);
}
for (m in 1:M){
    i[m] ~ normal(0,(1/alpha_i) * I);
}
}
"""

But I got a error:No matches for: row vector ~ normal(int, matrix) in this line of code:

for (n in 1: N){
    u[n] ~ normal(0,(1/alpha_u) * I);
}

Where I is a unit matrix. So the product of (1/alpha_u) * I is also a matrix. But stan just accept vector or real value as variance. I wonder how to transform it to a vector or a single value.

Thank you in advance!

Upvotes: 0

Views: 93

Answers (1)

Ben Goodrich
Ben Goodrich

Reputation: 4980

It appears as if you intend to be using the multivariate normal density, in which case the function you are looking for is multi_normal rather than normal. However, the multi_normal function needs a mean vector as its first argument, so you would need to call it as u[n] ~ multi_normal(rep_vector(0, K), (1/alpha_u) * I); In addition, there is no need to create I in the transformed parameters block when it should be created in the transformed data block.

All that said, you should never use the multi_normal density in Stan when the variance-covariance matrix is diagonal. Under this assumption, the multivariate normal distribution factors into a product of univariate normal distributions, so you would be better off doing u[n] ~ normal(0, 1 / alpha_u);

Upvotes: 1

Related Questions