Stochastic Gradient Descent outperforms Gradient Descent in recovering a high-dimensional signal in a glassy energy landscape.
Persia Jana KamaliPierfrancesco UrbaniPublished in: CoRR (2023)
Keyphrases
- stochastic gradient descent
- high dimensional
- energy landscape
- loss function
- least squares
- matrix factorization
- step size
- alternating least squares
- random forests
- simulated annealing
- coarse grained
- multiple kernel learning
- online algorithms
- dimensionality reduction
- weight vector
- low dimensional
- high dimensional data
- importance sampling
- feature space
- support vector machine
- neural network
- high level
- regularization parameter
- linear combination
- fine grained
- graph cuts
- particle swarm optimization
- probabilistic model
- cost function
- lower bound
- computer vision
- learning algorithm