Han et al. (2015) used a method of iterative pruning to reduce their network to only 10% of its original size with no loss of accuracy by removing weights with very low values, since these changed very little. As someone new to the machine learning area, why wouldn't you do this (unless your network is … Read more
Han et al. (2015) used a method of iterative pruning to reduce their network to only 10% of its original size with no loss of accuracy by removing weights with very low values, since these changed very little. As someone new to the machine learning area, why wouldn't you do this (unless your network is … Read more
I'm running an XGBRegressor for parameter optimization as below: xgb_model = xgb.XGBRegressor(nthread=10) clf = pipeline.Pipeline([('xgb', xgb_model)]) param_grid = {'xgb__max_depth': [1,2,3], 'xgb__learning_rate': [0.2,0.3,0.4], 'xgb__n_estimators': [40,45,50] } model = grid_search.GridSearchCV(estimator=clf, param_grid=param_grid, verbose=10, n_jobs=1, iid=True, refit=True, cv=10) model.fit(X, Y_log) I see few 'scores' listed (can be +/-) in the console out of which the 'Best score' is chosen. … Read more
I am estimating a simple AR(1) process by the ML approach. I also wish to compute the Quasi MLE standard errors, which is given by the sandwich form of the Hessian and the Score (see for example the last slide here) So, I start by just specifying the (conditional) log likelihood for the (gaussian) AR(1) … Read more
I am estimating a simple AR(1) process by the ML approach. I also wish to compute the Quasi MLE standard errors, which is given by the sandwich form of the Hessian and the Score (see for example the last slide here) So, I start by just specifying the (conditional) log likelihood for the (gaussian) AR(1) … Read more
I'm looking for advice on how I should approach a specific problem. I have about 1000 shops that I have to assign to about 20 different supply centers out of a possible 28, and I'm trying to pick the best 20 (or less) centers that minimizes the overall distance between shops and supply centers. I … Read more
I'm looking for advice on how I should approach a specific problem. I have about 1000 shops that I have to assign to about 20 different supply centers out of a possible 28, and I'm trying to pick the best 20 (or less) centers that minimizes the overall distance between shops and supply centers. I … Read more
Following Barto and Sutton's "Reinforcement Learning: An Introduction", I am having trouble rigorously proving the Bellman Optimality Equation for finite MDPs. Namely, why does $v_*(s) = maxlimits_{a in A(s)} q_{pi_*}(s, a)$? My attempt to see this is true: Let $v_* := maxlimits_{a in A(s)} q_{pi_*}(s, a)$ $v_*(s) = sumlimits_{a in A(s)} pi_*(a | s) q_{pi_*}(s, … Read more
Following Barto and Sutton's "Reinforcement Learning: An Introduction", I am having trouble rigorously proving the Bellman Optimality Equation for finite MDPs. Namely, why does $v_*(s) = maxlimits_{a in A(s)} q_{pi_*}(s, a)$? My attempt to see this is true: Let $v_* := maxlimits_{a in A(s)} q_{pi_*}(s, a)$ $v_*(s) = sumlimits_{a in A(s)} pi_*(a | s) q_{pi_*}(s, … Read more
I'm using random search for hyper-parameter optimization of a machine learning pipeline. For example, for the C and gamma parameter it is recommended to use logarithmically spaced values. Why should I use such values? For example, if I use logarithmic spaced values from $2^{-5}$ to $2^{15}$, then there will be many more values near to … Read more