regdiff 3.6.1__py3-none-any.whl

regdiff/__init__.py

@@ -0,0 +1 @@
+from .regdiff import finite_diff, matrix_anti_diff, matrix_diff, reg_diff  # noqa: F401

regdiff/regdiff.py

@@ -0,0 +1,240 @@
+import numpy as np
+import numpy.linalg as npl
+import scipy as sp
+
+# NOTE - all matrix/vector operations are called using np.matvel/np.subtract, etc. Only scalars should multiply ndarrays.
+
+
+def finite_diff(x):
+    """take the finite difference of the input vector. Performs no validation.
+    :param np.ndarray(x): data to take the finite difference of -- 1-dimensional array of length n
+
+    :return: diff(x) -- 1 dimensional array of length n-1
+    """
+    return np.subtract(x[1:], x[:-1])
+
+
+def matrix_diff(n, dx=1.0):
+    """return the matrix representations of the first-order central finite first difference, first-order central second
+    finite difference, and simple finite difference for a specified size and delta x increment.
+    :param n: the size of the requested matrix
+    :param dx: the time step
+
+    :return:    diff -- first order central finite first difference matrix of size n with free boundary conditions
+                diff2 -- first order central finite second difference matrix of size n with free boundary conditions
+                d -- simple finite differnceing matrix (nxn-1)
+    """
+    # first central difference
+    mul_step = 1 / (2 * dx)
+    c = mul_step * np.ones(n - 1)
+    diff = sp.sparse.diags_array([-c, c], offsets=[-1, 1]).toarray()
+    diff[0, 0:3] = mul_step * np.asarray([-3, 4, -1])
+    diff[-1, -3:] = mul_step * np.asarray([1, -4, 3])
+
+    # second central difference
+    mul_step = 1 / (dx**2)
+    c = mul_step * np.ones(n)
+    diff2 = sp.sparse.diags_array(
+        [c[0:-1], -2 * c, c[0:-1]], offsets=[-1, 0, 1]
+    ).toarray()
+    diff2[0, 0:4] = mul_step * np.asarray([2, -5, 4, -1])
+    diff2[-1, -4:] = mul_step * np.asarray([-1, 4, -5, 2])
+
+    # first order differencing matrix
+    mul_step = 1 / dx
+    c = mul_step * np.ones(n)
+    d = sp.sparse.diags_array([-c, c], offsets=[0, 1], shape=[n - 1, n]).toarray()
+
+    return diff, diff2, d
+
+
+def matrix_anti_diff(n, dx=1.0):
+    """return the matrix representation of the antidifferentiation operator for an input of size n
+    :param n: the size of the requested matrix
+    :param dx: the time step
+
+    :return:    a -- cumulative trapezoidal integration matrix that when multiplied by a vector yields the cumulative integral (without the first index) (nxn-1)
+                a2 -- cumulative trapezoidal double integration matrix that when multiplied by a vector yields the cumulative double integral (without the first index) (nxn-1)
+    """
+    c = 0.5 * dx * np.ones(n)
+    a = np.add(dx * np.tri(n, k=-1), np.diag(c))
+    a[:, 0] = 0.5 * dx
+    a[0, 0] = 0
+
+    a2 = np.matmul(a, a)
+
+    a = a[1:, :]
+    a2 = a2[1:, :]
+
+    return a, a2
+
+
+def reg_diff(
+    raw_data,
+    alpha,
+    alpha2=None,
+    dx=1,
+    u0=None,
+    min_iter=5,
+    max_iter=100000,
+    delta_cost_tol=1e-6,
+    delta_norm_tol=1e-6,
+    ep=1e-8,
+    diag=False,
+):
+    """differentation function that takes a vector of position data and returns the first and second derivatives, evaluated using TV-regularized differentiation.
+    :param data: the input data
+    :param alpha: the regularization coefficient for the first derivative
+    :param alpha2: the regularization coeffiient for the second derivative. If left blank only the first derivative will be calculated.
+    :param dx: the time step for the data
+    :param u0: an initial guess for the derivative(s) as an ndarray or list of ndarrays
+    :param min_iter: minimal number of iterations to calculate before returning. Default value 5.
+    :param max_iter: maximum number of iterations to calculate before returning. Default value is 1000000.
+    :param delta_cost_tol: relative delta cost to reach iteratively before returning (applies to both derivatives). Default value is 1e-6.
+    :param delta_norm_tol: relative delta norm of step to reach iteratively before returning (applies to both derivatives). Default value is 1e-6.
+    :param ep: epsilon to use to smooth the absolute value function. Devault value is 1e-8
+    :param diag: if True, prints diagnostic information during execution. Default False.
+
+    :return:    u -- the first derivative
+                v -- the second derivative
+    """
+    data = np.asarray(raw_data, copy=True)
+    n = len(data)
+    data_mean = data.mean()
+    data_st_dev = data.std()
+    data = (data - data_mean) / data_st_dev
+
+    if alpha2:
+        second_deriv = True
+    else:
+        second_deriv = False
+
+    diff, diff2, d = matrix_diff(n)
+    d_tp = np.transpose(d)
+
+    a, a2 = matrix_anti_diff(n)
+    a_tp = np.transpose(a)
+    a_tp_a = np.matmul(a_tp, a)
+    a_tp_a2 = np.matmul(a_tp, a2)
+
+    # bunch of code for handling an input u0
+    u = None
+    v = None
+    if u0 is not None:
+        if isinstance(u0, list):
+            if len(u0) == n:
+                u = np.asarray(u0, copy=True) * dx / data_st_dev
+            elif len(u0) == 2 * n:
+                u = np.asarray(u0[:n], copy=True) * dx / data_st_dev
+                v = np.asarray(u0[n:], copy=True) * (dx**2) / data_st_dev
+            elif len(u0) == 2 and len(u0[0]) == n and len(u0[1]) == n:
+                u = np.asarray(u0[0], copy=True) * dx / data_st_dev
+                v = np.asarray(u0[1], copy=True) * (dx**2) / data_st_dev
+        elif isinstance(u0, np.ndarray):
+            if u0.shape[0] == 2 and u0.shape[1] == n:
+                u = np.asarray(u0[0, :], copy=True) * dx / data_st_dev
+                v = np.asarray(u0[1, :], copy=True) * (dx**2) / data_st_dev
+            elif u0.shape[1] == 2 and u0.shape[0] == n:
+                u = np.asarray(u0[:, 0], copy=True) * dx / data_st_dev
+                v = np.asarray(u0[:, 1], copy=True) * (dx**2) / data_st_dev
+    if not isinstance(u, np.ndarray):
+        u = np.matvec(diff, data)
+    if not isinstance(v, np.ndarray) and second_deriv:
+        v = np.matvec(diff2, data)
+
+    # start calculating u
+    ofst_1 = data[0]
+    data_comp = data[1:] - ofst_1
+    a_tp_b = np.matvec(a_tp, data_comp)
+    if diag:
+        u_cost = [
+            npl.norm(np.subtract(sp.integrate.cumulative_trapezoid(u), data_comp))
+            + alpha * npl.norm(finite_diff(u), ord=1)
+        ]
+    prev_cost = None
+    for iter in range(max_iter):
+        c = np.reciprocal(np.sqrt(np.power(finite_diff(u), 2) + ep))
+        En = np.diag(c)
+        Ln = np.matmul(np.matmul(d_tp, En), d)
+        gn = np.add(np.subtract(np.matvec(a_tp_a, u), a_tp_b), alpha * np.matvec(Ln, u))
+        Hn = np.add(a_tp_a, alpha * Ln)
+        s = npl.solve(Hn, gn)
+        u = np.subtract(u, s)
+        cost_n = npl.norm(
+            np.subtract(sp.integrate.cumulative_trapezoid(u), data_comp)
+        ) + alpha * npl.norm(finite_diff(u), ord=1)
+        if diag:
+            print(
+                f"u, iteration {iter + 1}: cost: {cost_n}, relative change: {npl.norm(s) / npl.norm(u)}"
+            )
+            u_cost.append(cost_n)
+        if iter >= min_iter and npl.norm(s) / npl.norm(u) < delta_norm_tol:
+            if diag:
+                print("u, delta norm tolerance reached")
+            break
+        if (
+            prev_cost
+            and iter >= min_iter
+            and (prev_cost - cost_n) / prev_cost < delta_cost_tol
+        ):
+            if diag:
+                print("u, relative cost tolerance reached")
+            break
+        prev_cost = cost_n
+
+    if second_deriv:
+        ofst_2 = u[0]
+        data_comp = np.subtract(data[1:] - ofst_1, ofst_2 * np.arange(1, n))
+        a_tp_b = np.matvec(a_tp, data_comp)
+        if diag:
+            v_cost = [
+                npl.norm(
+                    np.subtract(
+                        sp.integrate.cumulative_trapezoid(
+                            sp.integrate.cumulative_trapezoid(v, initial=0)
+                        ),
+                        data_comp,
+                    )
+                )
+                + alpha2 * npl.norm(finite_diff(v), ord=1)
+            ]
+        prev_cost = None
+        for iter in range(max_iter):
+            c = np.reciprocal(np.sqrt(np.power(finite_diff(v), 2) + ep))
+            En = np.diag(c)
+            Ln = np.matmul(np.matmul(d_tp, En), d)
+            gn = np.add(
+                np.subtract(np.matvec(a_tp_a2, v), a_tp_b), alpha2 * np.matvec(Ln, v)
+            )
+            Hn = np.add(a_tp_a2, alpha2 * Ln)
+            s = npl.solve(Hn, gn)
+            v = np.subtract(v, s)
+            cost_n = npl.norm(
+                np.subtract(
+                    sp.integrate.cumulative_trapezoid(
+                        sp.integrate.cumulative_trapezoid(v, initial=0)
+                    ),
+                    data_comp,
+                )
+            ) + alpha2 * npl.norm(finite_diff(v), ord=1)
+            if diag:
+                print(
+                    f"v, iteration {iter + 1}: cost: {cost_n}, relative change: {npl.norm(s) / npl.norm(v)}"
+                )
+                v_cost.append(cost_n)
+            if iter >= min_iter and npl.norm(s) / npl.norm(v) < delta_norm_tol:
+                if diag:
+                    print("v, delta norm tolerance reached")
+                break
+            if (
+                prev_cost
+                and iter >= min_iter
+                and (prev_cost - cost_n) / prev_cost < delta_cost_tol
+            ):
+                if diag:
+                    print("v, relative cost tolerance reached")
+                break
+            prev_cost = cost_n
+        v = v * data_st_dev / (dx**2)
+    u = u * data_st_dev / dx
+    return u, v

regdiff-3.6.1.dist-info/METADATA

@@ -0,0 +1,7 @@
+Metadata-Version: 2.3
+Name: regdiff
+Version: 3.6.1
+Summary: Regularized differentiation of time-series as numpy ndarrays
+Requires-Dist: numpy>=2.5.4
+Requires-Dist: scipy>=1.18.1
+Requires-Python: >=3.14

regdiff-3.6.1.dist-info/RECORD

@@ -0,0 +1,6 @@
+regdiff/__init__.py,sha256=_RqYKdJ6fYXBMBVcO8SuCV8uR5sWF1SIf4rtuuIRo5w,88
+regdiff/regdiff.py,sha256=KvOXaP2lSY_YwLGdBGtQMbWZ34_X2NGWz4QM6nwTnXo,9647
+regdiff-3.6.1.dist-info/WHEEL,sha256=_d8F1e7SqtoW6CDj4Gi8lFC26a_7I17R7zPLCKTp4Fg,81
+regdiff-3.6.1.dist-info/entry_points.txt,sha256=wQIoF7cZJ4p4m9kOMpU8WwSZioY6AfiPSTBt3ZFwwR8,42
+regdiff-3.6.1.dist-info/METADATA,sha256=apkvwLLp0qfIKVRGZ_PjOpdgm371ohPYWRK3cuFfK2I,202
+regdiff-3.6.1.dist-info/RECORD,,

regdiff-3.6.1.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: uv 0.12.15
+Root-Is-Purelib: true
+Tag: py3-none-any

regdiff-3.6.1.dist-info/entry_points.txt

@@ -0,0 +1,3 @@
+[console_scripts]
+regdiff = regdiff:main
+