Merge pull request #387 from QuantEcon/interp_fixes

Use np.interp instead of interpolation.interp

Author: jstac

lectures/_static/lecture_specific/optgrowth/bellman_operator.py

@@ -15,15 +15,6 @@ kernelspec:
 :depth: 2
 ```
 
-In addition to what's in Anaconda, this lecture will require the following library:
-
-```{code-cell} ipython
----
-tags: [hide-output]
----
-!pip install interpolation
-```
-
 ## Overview
 
 In this lecture we continue the study of {doc}`the cake eating problem <cake_eating_problem>`.
@@ -47,9 +38,7 @@ We will use the following imports:
 
 ```{code-cell} ipython
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
-from interpolation import interp
 from scipy.optimize import minimize_scalar, bisect
 ```
 
@@ -211,7 +200,7 @@ class CakeEating:
         """
 
         u, β = self.u, self.β
-        v = lambda x: interp(self.x_grid, v_array, x)
+        v = lambda x: np.interp(x, self.x_grid, v_array)
 
         return u(c) + β * v(x - c)
 ```
@@ -533,7 +522,7 @@ class OptimalGrowth(CakeEating):
         """
 
         u, β, α = self.u, self.β, self.α
-        v = lambda x: interp(self.x_grid, v_array, x)
+        v = lambda x: np.interp(x, self.x_grid, v_array)
 
         return u(c) + β * v((x - c)**α)
 ```
@@ -609,7 +598,7 @@ def K(σ_array, ce):
     u_prime, β, x_grid = ce.u_prime, ce.β, ce.x_grid
     σ_new = np.empty_like(σ_array)
 
-    σ = lambda x: interp(x_grid, σ_array, x)
+    σ = lambda x: np.interp(x, x_grid, σ_array)
 
     def euler_diff(c, x):
         return u_prime(c) - β * u_prime(σ(x - c))

lectures/cake_eating_numerical.md

@@ -30,7 +30,6 @@ In addition to what's in Anaconda, this lecture will need the following librarie
 tags: [hide-output]
 ---
 !pip install quantecon
-!pip install interpolation
 ```
 
 ## Overview
@@ -62,9 +61,7 @@ Let's start with some imports:
 
 ```{code-cell} ipython
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
-from interpolation import interp
 from quantecon.optimize import brentq
 from numba import njit
 ```
@@ -301,7 +298,7 @@ def euler_diff(c, σ, y, og):
     f, f_prime, u_prime = og.f, og.f_prime, og.u_prime
 
     # First turn σ into a function via interpolation
-    σ_func = lambda x: interp(grid, σ, x)
+    σ_func = lambda x: np.interp(x, grid, σ)
 
     # Now set up the function we need to find the root of.
     vals = u_prime(σ_func(f(y - c) * shocks)) * f_prime(y - c) * shocks

lectures/coleman_policy_iter.md

@@ -23,14 +23,6 @@ kernelspec:
 :depth: 2
 ```
 
-In addition to what's in Anaconda, this lecture will need the following libraries:
-
-```{code-cell} ipython
----
-tags: [hide-output]
----
-!pip install interpolation
-```
 
 ## Overview
 
@@ -51,9 +43,7 @@ Let's start with some standard imports:
 
 ```{code-cell} ipython
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
-from interpolation import interp
 from numba import njit
 ```
 
@@ -188,7 +178,7 @@ def K(σ_array, og):
     y = grid + σ_array  # y_i = k_i + c_i
 
     # Linear interpolation of policy using endogenous grid
-    σ = lambda x: interp(y, σ_array, x)
+    σ = lambda x: np.interp(x, y, σ_array)
 
     # Allocate memory for new consumption array
     c = np.empty_like(grid)

lectures/egm_policy_iter.md

@@ -30,7 +30,6 @@ In addition to what's in Anaconda, this lecture will need the following librarie
 tags: [hide-output]
 ---
 !pip install quantecon
-!pip install interpolation
 ```
 
 ## Overview
@@ -60,10 +59,8 @@ We'll need the following imports:
 
 ```{code-cell} ipython
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
 from quantecon.optimize import brentq
-from interpolation import interp
 from numba import njit, float64
 from numba.experimental import jitclass
 from quantecon import MarkovChain
@@ -437,7 +434,7 @@ def euler_diff(c, a, z, σ_vals, ifp):
 
     # Convert policy into a function by linear interpolation
     def σ(a, z):
-        return interp(asset_grid, σ_vals[:, z], a)
+        return np.interp(a, asset_grid, σ_vals[:, z])
 
     # Calculate the expectation conditional on current z
     expect = 0.0
@@ -663,7 +660,7 @@ def compute_asset_series(ifp, T=500_000, seed=1234):
 
     # Solve for the optimal policy
     σ_star = solve_model_time_iter(ifp, σ_init, verbose=False)
-    σ = lambda a, z: interp(ifp.asset_grid, σ_star[:, z], a)
+    σ = lambda a, z: np.interp(a, ifp.asset_grid, σ_star[:, z])
 
     # Simulate the exogeneous state process
     mc = MarkovChain(P)

lectures/ifp.md

@@ -30,7 +30,6 @@ In addition to what's in Anaconda, this lecture will need the following librarie
 tags: [hide-output]
 ---
 !pip install quantecon
-!pip install interpolation
 ```
 
 ## Overview
@@ -57,9 +56,7 @@ We require the following imports:
 
 ```{code-cell} ipython
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
-from interpolation import interp
 from numba import njit, float64
 from numba.experimental import jitclass
 from quantecon import MarkovChain
@@ -436,7 +433,7 @@ def K(a_in, σ_in, ifp):
     n = len(P)
 
     # Create consumption function by linear interpolation
-    σ = lambda a, z: interp(a_in[:, z], σ_in[:, z], a)
+    σ = lambda a, z: np.interp(a, a_in[:, z], σ_in[:, z])
 
     # Allocate memory
     σ_out = np.empty_like(σ_in)
@@ -636,7 +633,7 @@ def compute_asset_series(ifp, a_star, σ_star, z_seq, T=500_000):
     """
 
     # Create consumption function by linear interpolation
-    σ = lambda a, z: interp(a_star[:, z], σ_star[:, z], a)
+    σ = lambda a, z: np.interp(a, a_star[:, z], σ_star[:, z])
 
     # Simulate the asset path
     a = np.zeros(T+1)

lectures/ifp_advanced.md

@@ -27,15 +27,6 @@ kernelspec:
 :depth: 2
 ```
 
-In addition to what's in Anaconda, this lecture will need the following libraries:
-
-```{code-cell} ipython
----
-tags: [hide-output]
----
-!pip install interpolation
-```
-
 ## Overview
 
 In this section, we solve a simple on-the-job search model
@@ -46,10 +37,8 @@ Let's start with some imports:
 
 ```{code-cell} ipython
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
 import scipy.stats as stats
-from interpolation import interp
 from numba import njit, prange
 ```
 
@@ -269,7 +258,7 @@ def operator_factory(jv, parallel_flag=True):
     @njit
     def state_action_values(z, x, v):
         s, ϕ = z
-        v_func = lambda x: interp(x_grid, v, x)
+        v_func = lambda x: np.interp(x, x_grid, v)
 
         integral = 0
         for m in range(mc_size):
@@ -460,8 +449,8 @@ v_star = solve_model(jv, verbose=False)
 s_policy, ϕ_policy = get_greedy(v_star)
 
 # Turn the policy function arrays into actual functions
-s = lambda y: interp(x_grid, s_policy, y)
-ϕ = lambda y: interp(x_grid, ϕ_policy, y)
+s = lambda y: np.interp(y, x_grid, s_policy)
+ϕ = lambda y: np.interp(y, x_grid, ϕ_policy)
 
 def h(x, b, u):
     return (1 - b) * g(x, ϕ(x)) + b * max(g(x, ϕ(x)), u)

lectures/jv.md

@@ -30,7 +30,6 @@ In addition to what's in Anaconda, this lecture will need the following librarie
 tags: [hide-output]
 ---
 !pip install quantecon
-!pip install interpolation
 ```
 
 ## Overview
@@ -48,10 +47,8 @@ We will use the following imports:
 
 ```{code-cell} ipython3
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
 import quantecon as qe
-from interpolation import interp
 from numpy.random import randn
 from numba import njit, prange, float64
 from numba.experimental import jitclass
@@ -245,7 +242,7 @@ def Q(js, f_in, f_out):
         for m in range(M):
             e1, e2 = js.e_draws[:, m]
             z_next = d + ρ * z + σ * e1
-            go_val = interp(js.z_grid, f_in, z_next)     # f(z')
+            go_val = np.interp(z_next, js.z_grid, f_in)  # f(z')
             y_next = np.exp(μ + s * e2)                  # y' draw
             w_next = np.exp(z_next) + y_next             # w' draw
             stop_val = np.log(w_next) / (1 - β)
@@ -353,7 +350,7 @@ def compute_unemployment_duration(js, seed=1234):
 
     @njit
     def f_star_function(z):
-        return interp(z_grid, f_star, z)
+        return np.interp(z, z_grid, f_star)
 
     @njit
     def draw_tau(t_max=10_000):

lectures/mccall_correlated.md

@@ -23,14 +23,6 @@ kernelspec:
 :depth: 2
 ```
 
-In addition to what's in Anaconda, this lecture will need the following libraries:
-
-```{code-cell} ipython
----
-tags: [hide-output]
----
-!pip install interpolation
-```
 
 ## Overview
 
@@ -58,9 +50,7 @@ We will use the following imports:
 
 ```{code-cell} ipython3
 import matplotlib.pyplot as plt
-plt.rcParams["figure.figsize"] = (11, 5)  #set default figure size
 import numpy as np
-from interpolation import interp
 from numba import njit, float64
 from numba.experimental import jitclass
 ```
@@ -155,7 +145,7 @@ This method
    {cite}`gordon1995stable` or {cite}`stachurski2008continuous`) and
 1. preserves useful shape properties such as monotonicity and concavity/convexity.
 
-Linear interpolation will be implemented using a JIT-aware Python interpolation library called [interpolation.py](https://github.com/EconForge/interpolation.py).
+Linear interpolation will be implemented using [numpy.interp](https://numpy.org/doc/stable/reference/generated/numpy.interp.html).
 
 The next figure illustrates piecewise linear interpolation of an arbitrary
 function on grid points $0, 0.2, 0.4, 0.6, 0.8, 1$.
@@ -169,7 +159,7 @@ c_grid = np.linspace(0, 1, 6)
 f_grid = np.linspace(0, 1, 150)
 
 def Af(x):
-    return interp(c_grid, f(c_grid), x)
+    return np.interp(x, c_grid, f(c_grid))
 
 fig, ax = plt.subplots()
 
@@ -238,7 +228,7 @@ class McCallModelContinuous:
         u = lambda x: np.log(x)
 
         # Interpolate array represented value function
-        vf = lambda x: interp(w, v, x)
+        vf = lambda x: np.interp(x, w, v)
 
         # Update d using Monte Carlo to evaluate integral
         d_new = np.mean(np.maximum(vf(self.w_draws), u(c) + β * d))