During the training of Knowledge Tracing Machines, the negative log-likelihood function is minimized: $NLL(p(X), y) = \sum_{i=1}^S y_i \log p(x_i) + (1-y_i) \log (1-p(x_i))$ Here, $y_i$ represents the outcomes and $x_i$ represents the inputs. Biases and embeddings follow hyperpriors: $w_k \sim N(\mu, 1/\lambda)$ and $v_{kf} \sim N(\mu, 1/\lambda)$. The parameters $\mu$ and $\lambda$ follow hyperpriors: mu sim N(0, 1) and lambda sim Gamma(0, 1). Due to the use of these hyperpriors, fine-tuning is not required.