Merge pull request #44 from ISISComputingGroup/fixes_linear_fitting

Fixes issues found with linear fitting guess function

Author: Tom-Willemsen

doc/fitting/fitting.md

@@ -111,9 +111,9 @@ See the following example on how to define these.
 
 import lmfit
 
-def model(x: float, m: float, c: float) -> float:
+def model(x: float, c1: float, c0: float) -> float:
 
-    return m * x + c # y = mx + c
+    return c1 * x + c0 # y = mx + c
 
 def guess(x: npt.NDArray[np.float64], y: npt.NDArray[np.float64]) -> dict[str, lmfit.Parameter]:
 
@@ -125,12 +125,12 @@ def guess(x: npt.NDArray[np.float64], y: npt.NDArray[np.float64]) -> dict[str, l
     numerator = sum(x * y) - sum(x) * sum(y)
     denominator = sum(x**2) - sum(x) ** 2
 
-    m = numerator / denominator
-    c = (sum(y) - m * sum(x)) / len(x)
+    c1 = numerator / denominator
+    c0 = (sum(y) - c1 * sum(x)) / len(x)
 
     init_guess = {
-        "m": lmfit.Parameter("m", m), # gradient
-        "c": lmfit.Parameter("c", c), # y - intercept
+        "c1": lmfit.Parameter("c1", c1), # gradient
+        "c0": lmfit.Parameter("c0", c0), # y - intercept
     }
 
     return init_guess
@@ -155,9 +155,9 @@ This means that aslong as the parameters returned from the guess function match
 import lmfit
 from ibex_bluesky_core.callbacks.fitting.fitting_utils import Linear
 
-def different_model(x: float, m: float, c: float) -> float:
+def different_model(x: float, c1: float, c0: float) -> float:
 
-    return m * x + c ** 2 # y = mx + (c ** 2)
+    return c1 * x + c0 ** 2 # y = mx + (c ** 2)
 
 
 fit_method = FitMethod(different_model, Linear.guess())
@@ -176,11 +176,11 @@ from ibex_bluesky_core.callbacks.fitting.fitting_utils import Linear
 # This Guessing. function isn't very good because it's return values don't change on the data already collected in the Bluesky run
 # It always guesses that the linear function is y = x
 
-def different_guess(x: float, m: float, c: float) -> float:
+def different_guess(x: float, c1: float, c0: float) -> float:
 
     init_guess = {
-        "m": lmfit.Parameter("m", 1), # gradient
-        "c": lmfit.Parameter("c", 0), # y - intercept
+        "c1": lmfit.Parameter("c1", 1), # gradient
+        "c0": lmfit.Parameter("c0", 0), # y - intercept
     }
 
     return init_guess

doc/fitting/standard_fits.md

@@ -2,20 +2,20 @@
 
 ## Linear
 
-- `m` - Gradient
-- `c` - (y) Intercept
+- `c1` - Gradient
+- `c0` - (y) Intercept
 
 ```{math}
-y = mx + c
+y = c_1x + c_0
 ```
 
 ## Polynomial
 
-- `a` ... `d` - Polynomial coefficients
+- `cn` ... `c0` - Polynomial coefficients
 
 For a polynomial degree `n`:
 ```{math}
-y = ax^n + bx^n-1 + ... + cx^1 + d 
+y = c_{n}x^n + c_{n-1}x^n-1 + ... + c_1 * x^1 + c_0 
 ```
 
 ## Gaussian

src/ibex_bluesky_core/callbacks/fitting/fitting_utils.py

@@ -179,8 +179,8 @@ class Linear(Fit):
     def model(cls, *args: int) -> lmfit.Model:
         """Linear Model."""
 
-        def model(x: npt.NDArray[np.float64], m: float, c: float) -> npt.NDArray[np.float64]:
-            return m * x + c
+        def model(x: npt.NDArray[np.float64], c1: float, c0: float) -> npt.NDArray[np.float64]:
+            return c1 * x + c0
 
         return lmfit.Model(model)
 
@@ -189,25 +189,8 @@ def guess(
         cls, *args: int
     ) -> Callable[[npt.NDArray[np.float64], npt.NDArray[np.float64]], dict[str, lmfit.Parameter]]:
         """Linear Guessing."""
-
-        def guess(
-            x: npt.NDArray[np.float64], y: npt.NDArray[np.float64]
-        ) -> dict[str, lmfit.Parameter]:
-            # Linear Regression
-            numerator = sum(x * y) - sum(x) * sum(y)
-            denominator = sum(x**2) - sum(x) ** 2
-
-            m = numerator / denominator
-            c = (sum(y) - m * sum(x)) / len(x)
-
-            init_guess = {
-                "m": lmfit.Parameter("m", m),
-                "c": lmfit.Parameter("c", c),
-            }
-
-            return init_guess
-
-        return guess
+        return Polynomial.guess(1)
+        # Uses polynomial guessing with a degree of 1
 
 
 class Polynomial(Fit):

tests/callbacks/fitting/test_fitting_methods.py

@@ -219,7 +219,7 @@ def test_linear_model(self):
             grad = 3
             y_intercept = 2
 
-            outp = Linear.model().func(x, m=grad, c=y_intercept)
+            outp = Linear.model().func(x, c1=grad, c0=y_intercept)
 
             # Check for constant gradient of grad
             outp_m = np.diff(outp)
@@ -234,19 +234,26 @@ def test_gradient_guess(self):
             y = np.array([-1.0, 0.0, 1.0])
             outp = Linear.guess()(x, y)
 
-            assert pytest.approx(outp["m"]) == 1.0
+            assert pytest.approx(outp["c1"]) == 1.0
 
             y = np.array([-2.0, 0.0, 2.0])
             outp1 = Linear.guess()(x, y)
             # check with a graph with steeper gradient
-            assert outp["m"] < outp1["m"]
+            assert outp["c1"] < outp1["c1"]
 
         def test_y_intercept_guess(self):
             x = np.array([-1.0, 0.0, 1.0])
             y = np.array([-1.0, 0.0, 1.0])
             outp = Linear.guess()(x, y)
 
-            assert pytest.approx(outp["c"]) == 0.0
+            assert pytest.approx(outp["c0"]) == 0.0
+
+        def test_zero_gradient_guess(self):
+            x = np.array([-1.0, 0.0, 1.0])
+            y = np.array([0.0, 0.0, 0.0])
+            outp = Linear.guess()(x, y)
+
+            assert pytest.approx(outp["c1"]) == 0.0
 
 
 class TestPolynomial: