reorganize GLM penalties, added apdt lasso

Author: idc9

Readme.md

@@ -115,7 +115,7 @@ EstCV(estimator=est).fit(X, y)
 
 # Quantile regression with your favorite optimization algorithm
 # you can easily provide your own optimization algorithm to be the backend solver
-from ya_glm.backends.quantile_lp.glm_solver import solve_glm # Linear Programming formulation of quantile regression
+from ya_glm.backends.quantile_lp.glm_solver import solve_glm
 
 Est, EstCV = get_pen_glm(loss_func='quantile',
                          penalty='adpt_lasso',

docs/concave_penalties_compare_initializations.ipynb

@@ -16,7 +16,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 3,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -25,7 +25,7 @@
     "\n",
     "from ya_glm.toy_data import sample_sparse_lin_reg\n",
     "from ya_glm.backends.fista.LinearRegression import FcpLLA, FcpLLACV,\\\n",
-    "    LassoCV, RidgeCV, LassoENetCV"
+    "    LassoCV, RidgeCV, ENetCV"
    ]
   },
   {
@@ -37,7 +37,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 4,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -49,7 +49,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 5,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -84,7 +84,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": 6,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,48 +96,24 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "CPU times: user 152 ms, sys: 81.5 ms, total: 233 ms\n",
-      "Wall time: 3.7 s\n",
+      "CPU times: user 142 ms, sys: 75.6 ms, total: 218 ms\n",
+      "Wall time: 3.16 s\n",
       "Lasso L2 to truth 0.5399115512254988 \n",
       "\n",
-      "CPU times: user 246 ms, sys: 84 ms, total: 330 ms\n",
-      "Wall time: 1.08 s\n",
+      "CPU times: user 257 ms, sys: 117 ms, total: 374 ms\n",
+      "Wall time: 836 ms\n",
       "Ridge L2 to truth 1.4667162260738444 \n",
       "\n",
-      "CPU times: user 1.02 s, sys: 77.9 ms, total: 1.1 s\n",
-      "Wall time: 18.1 s\n",
+      "CPU times: user 823 ms, sys: 66.4 ms, total: 889 ms\n",
+      "Wall time: 13.9 s\n",
       "ENet L2 to truth 0.5399115512254988 \n",
-      "\n",
-      "CPU times: user 1.92 s, sys: 221 ms, total: 2.14 s\n",
-      "Wall time: 6.35 s\n",
-      "FCP, lasso init, one step L2 to truth 0.2934270519442212 \n",
-      "\n",
-      "CPU times: user 2.01 s, sys: 205 ms, total: 2.21 s\n",
-      "Wall time: 11.1 s\n",
-      "FCP, ridge init, one step L2 to truth 0.3433528089934003 \n",
-      "\n",
-      "CPU times: user 2.65 s, sys: 196 ms, total: 2.85 s\n",
-      "Wall time: 21.1 s\n",
-      "FCP, enet init, one step L2 to truth 0.2934270519442212 \n",
-      "\n",
-      "CPU times: user 2.59 s, sys: 218 ms, total: 2.81 s\n",
-      "Wall time: 18.4 s\n",
-      "FCP, lasso init, many steps L2 to truth 0.2964122567536647 \n",
-      "\n",
-      "CPU times: user 2.56 s, sys: 265 ms, total: 2.83 s\n",
-      "Wall time: 19.7 s\n",
-      "FCP, ridge init, many steps L2 to truth 0.2964122567536647 \n",
-      "\n",
-      "CPU times: user 2.85 s, sys: 201 ms, total: 3.05 s\n",
-      "Wall time: 28.5 s\n",
-      "FCP, enet init, many steps L2 to truth 0.2964122567536647 \n",
       "\n"
      ]
     }
@@ -157,7 +133,7 @@
     "      np.linalg.norm(ridge.best_estimator_.coef_ - coef), '\\n')\n",
     "\n",
     "# ElasticNet\n",
-    "%time enet = LassoENetCV(l1_ratio='tune', **cv_kws).fit(X, y)\n",
+    "%time enet = ENetCV(l1_ratio='tune', **cv_kws).fit(X, y)\n",
     "print('ENet L2 to truth',\n",
     "      np.linalg.norm(enet.best_estimator_.coef_ - coef), '\\n')\n",
     "\n",

docs/concave_penalty_two_stage_estimators.ipynb

@@ -53,7 +53,7 @@
     {
      "data": {
       "text/plain": [
-       "<function ya_glm.fcp.GlmFcp.GlmFcpFitLLA.__init__(self, lla_n_steps=1, lla_kws={}, fit_intercept=True, standardize=False, opt_kws={}, pen_val=1, pen_func='scad', pen_func_kws={}, init='default')>"
+       "<function ya_glm.pen_glms.GlmFcpLLA.GlmFcpLLA.__init__(self, pen_val=1, pen_func='scad', pen_func_kws={}, init='default', lla_n_steps=1, lla_kws={}, ridge_pen_val=None, ridge_weights=None, tikhonov=None, groups=None, fit_intercept=True, standardize=False, opt_kws={})>"
       ]
      },
      "execution_count": 3,
@@ -160,6 +160,13 @@
     "est.fit(X, y)\n",
     "est.coef_"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {