Merge pull request #165 from pfizer-opensource/development

LukasAdamowicz · web-flow · commit c88ddc3caaa0 · 2024-06-24T10:44:17.000-04:00
Version 0.16.3
diff --git a/docs/src/research.rst b/docs/src/research.rst
@@ -5,6 +5,7 @@ SKDH Research
 Peer Reviewed Journal Articles Using SKDH
 =========================================
 
+- N. Camerlingo et al., “Measuring gait parameters from a single chest-worn accelerometer in healthy individuals: a validation study,” Sci Rep, vol. 14, no. 1, p. 13897, Jun. 2024, doi: 10.1038/s41598-024-62330-6.
 - W. Lin et al., “Can Gait Characteristics Be Represented by Physical Activity Measured with Wrist-Worn Accelerometers?,” Sensors, vol. 23, no. 20, Art. no. 20, Jan. 2023, doi: 10.3390/s23208542.
 - W. Lin et al., "SciKit Digital Health Package for Accelerometry-Measured Physical Activity: Comparisons to Existing Solutions and Investigations of Age Effects in Healthy Adults", Front. Digit. Health, 26 Nov. 2023, doi: 10.3389/fdgth.2023.1321086
 - J. Di et al., “Monitoring Activity and Gait in Children (MAGIC) using digital health technologies,” Pediatr Res, Mar. 2024, doi: 10.1038/s41390-024-03147-x.
diff --git a/meson.build b/meson.build
@@ -1,7 +1,7 @@
 project(
     'scikit-digital-health',
     'c',
-    version: '0.16.2',
+    version: '0.16.3',
     license: 'MIT',
     meson_version: '>=1.1',
 )
diff --git a/pyproject.toml b/pyproject.toml
@@ -2,7 +2,7 @@
 requires = [
     "meson-python",
     "wheel",
-    "oldest-supported-numpy"
+    "numpy>=2.0.0rc1"
 ]
 build-backend = "mesonpy"
 
@@ -47,7 +47,7 @@ classifiers = [
 
 requires-python = ">=3.9"
 dependencies = [
-    "numpy>=1.17.2",
+    "numpy>=1.25.0",
     "scipy>=1.12.0",
     "pandas>=1.0.0",
     "lightgbm>=2.3.0",
@@ -65,7 +65,8 @@ dev = [
     'pytest',
     'coverage',
     'psutil',
-    'tables'
+    # 'tables',
+    "numpy>=2.0.0rc1"
 ]
 actions = [
     'flake8',
diff --git a/src/skdh/features/core.py b/src/skdh/features/core.py
@@ -11,7 +11,7 @@
 from warnings import warn
 
 from pandas import DataFrame
-from numpy import float_, asarray, zeros, sum, moveaxis
+from numpy import float64, asarray, zeros, sum, moveaxis
 
 
 __all__ = ["Bank"]
@@ -262,10 +262,10 @@ def compute(
         # standardize the input signal
         if isinstance(signal, DataFrame):
             columns = columns if columns is not None else signal.columns
-            x = signal[columns].values.astype(float_)
+            x = signal[columns].values.astype(float64)
         else:
             try:
-                x = asarray(signal, dtype=float_)
+                x = asarray(signal, dtype=float64)
             except ValueError as e:
                 raise ArrayConversionError("Error converting signal to ndarray") from e
 
@@ -285,7 +285,7 @@ def compute(
             x = moveaxis(x, axis, -1)
             # number of feats is 1 per
             n_feats = [1] * len(self)
-            feats = zeros((sum(n_feats),) + x.shape[:-1], dtype=float_)
+            feats = zeros((sum(n_feats),) + x.shape[:-1], dtype=float64)
         else:
             # move both the computation and index axis. do this in two steps to allow for undoing
             # just the index axis swap later. The index_axis has been adjusted appropriately
@@ -297,7 +297,7 @@ def compute(
             for ind in indices:
                 n_feats.append(get_n_feats(x.shape[0], ind))
 
-            feats = zeros((sum(n_feats),) + x.shape[1:-1], dtype=float_)
+            feats = zeros((sum(n_feats),) + x.shape[1:-1], dtype=float64)
 
         feat_i = 0  # keep track of where in the feature array we are
         for i, ft in enumerate(self._feats):
@@ -367,4 +367,4 @@ def compute(self, signal, fs=1.0, *, axis=-1):
             ndarray of the computed feature
         """
         # move the computation axis to the end
-        return moveaxis(asarray(signal, dtype=float_), axis, -1)
+        return moveaxis(asarray(signal, dtype=float64), axis, -1)
diff --git a/src/skdh/gait/__init__.py b/src/skdh/gait/__init__.py
@@ -280,7 +280,7 @@ def __init__(self):
 
         def _predict(self, dt, leg_length, gait, gait_aux):
             # initialize 1 value per bout
-            stepreg = zeros(len(gait_aux['accel']), dtype=float_)
+            stepreg = zeros(len(gait_aux['accel']), dtype=float64)
 
             for i, acc in enumerate(gait_aux['accel']):
                 lag = int(
diff --git a/src/skdh/gait/gait_metrics/base.py b/src/skdh/gait/gait_metrics/base.py
@@ -8,7 +8,7 @@
 import functools
 import logging
 
-from numpy import zeros, roll, full, nan, bool_, float_
+from numpy import zeros, roll, full, nan, bool_, float64
 
 
 def basic_asymmetry(f):
@@ -134,7 +134,7 @@ def predict(self, *, fs, leg_length, gait, gait_aux):
 
     def _predict_asymmetry(self, *, fs, leg_length, gait, gait_aux):
         asy_name = f"{self.k_} asymmetry"
-        gait[asy_name] = full(gait["IC"].size, nan, dtype=float_)
+        gait[asy_name] = full(gait["IC"].size, nan, dtype=float64)
 
         mask = self._get_mask(gait, 1)
         mask_ofst = roll(mask, 1)
@@ -143,7 +143,7 @@ def _predict_asymmetry(self, *, fs, leg_length, gait, gait_aux):
 
     def _predict_init(self, gait, init=True, offset=None):
         if init:
-            gait[self.k_] = full(gait["IC"].size, nan, dtype=float_)
+            gait[self.k_] = full(gait["IC"].size, nan, dtype=float64)
         if offset is not None:
             mask = self._get_mask(gait, offset)
             mask_ofst = roll(mask, offset)
diff --git a/src/skdh/gait/gait_metrics/gait_metrics.py b/src/skdh/gait/gait_metrics/gait_metrics.py
@@ -44,7 +44,7 @@
     argmin,
     abs,
     round,
-    float_,
+    float64,
     int_,
     fft,
     arange,
@@ -101,7 +101,7 @@ def _autocovariancefn(x, max_lag, biased=False, axis=0):
 
     shape = list(y.shape)
     shape[-1] = max_lag
-    ac = zeros(shape, dtype=float_)
+    ac = zeros(shape, dtype=float64)
 
     for i in range(min(max_lag, y.shape[-1] - 10)):
         ac[..., i] = autocorrelation(y, i, True)
@@ -796,7 +796,7 @@ def __init__(self):
         )
 
     def _predict(self, *, fs, leg_length, gait, gait_aux):
-        pci = zeros(len(gait_aux["accel"]), dtype=float_)
+        pci = zeros(len(gait_aux["accel"]), dtype=float64)
 
         phase = gait["step time"] / gait["stride time"]  # %, not degrees
         for i in range(len(gait_aux["accel"])):
@@ -869,7 +869,7 @@ def __init__(self):
         super().__init__("gait symmetry index", __name__, depends=[StrideTime])
 
     def _predict(self, *, fs, leg_length, gait, gait_aux):
-        gsi = zeros(len(gait_aux["accel"]), dtype=float_)
+        gsi = zeros(len(gait_aux["accel"]), dtype=float64)
 
         # setup acceleration filter if its possible to use
         if 0 < (2 * 10 / fs) < 1:
@@ -942,7 +942,7 @@ def __init__(self):
         super().__init__("step regularity - V", __name__, depends=[StepTime])
 
     def _predict(self, *, fs, leg_length, gait, gait_aux):
-        stepreg = full(len(gait_aux["accel"]), nan, dtype=float_)
+        stepreg = full(len(gait_aux["accel"]), nan, dtype=float64)
 
         for i in unique(gait_aux["inertial data i"]):
             acc = gait_aux["accel"][i]
@@ -1002,7 +1002,7 @@ def __init__(self):
         super().__init__("stride regularity - V", __name__, depends=[StrideTime])
 
     def _predict(self, *, fs, leg_length, gait, gait_aux):
-        stridereg = full(len(gait_aux["accel"]), nan, dtype=float_)
+        stridereg = full(len(gait_aux["accel"]), nan, dtype=float64)
 
         for i in unique(gait_aux["inertial data i"]):
             acc = gait_aux["accel"][i]
diff --git a/src/skdh/gait_old/__init__.py b/src/skdh/gait_old/__init__.py
@@ -174,7 +174,7 @@ def __init__(self):
 
         def _predict(self, dt, leg_length, gait, gait_aux):
             # initialize 1 value per bout
-            stepreg = zeros(len(gait_aux['accel']), dtype=float_)
+            stepreg = zeros(len(gait_aux['accel']), dtype=float64)
 
             for i, acc in enumerate(gait_aux['accel']):
                 lag = int(
diff --git a/src/skdh/gait_old/gait_endpoints/base.py b/src/skdh/gait_old/gait_endpoints/base.py
@@ -8,7 +8,7 @@
 import functools
 import logging
 
-from numpy import zeros, roll, full, nan, bool_, float_
+from numpy import zeros, roll, full, nan, bool_, float64
 
 
 def basic_asymmetry(f):
@@ -130,7 +130,7 @@ def predict(self, fs, leg_length, gait, gait_aux):
 
     def _predict_asymmetry(self, dt, leg_length, gait, gait_aux):
         asy_name = f"{self.k_} asymmetry"
-        gait[asy_name] = full(gait["IC"].size, nan, dtype=float_)
+        gait[asy_name] = full(gait["IC"].size, nan, dtype=float64)
 
         mask = self._get_mask(gait, 1)
         mask_ofst = roll(mask, 1)
@@ -139,7 +139,7 @@ def _predict_asymmetry(self, dt, leg_length, gait, gait_aux):
 
     def _predict_init(self, gait, init=True, offset=None):
         if init:
-            gait[self.k_] = full(gait["IC"].size, nan, dtype=float_)
+            gait[self.k_] = full(gait["IC"].size, nan, dtype=float64)
         if offset is not None:
             mask = self._get_mask(gait, offset)
             mask_ofst = roll(mask, offset)
diff --git a/src/skdh/gait_old/gait_endpoints/gait_endpoints.py b/src/skdh/gait_old/gait_endpoints/gait_endpoints.py
@@ -44,7 +44,7 @@
     argmin,
     abs,
     round,
-    float_,
+    float64,
     int_,
     fft,
     arange,
@@ -97,7 +97,7 @@ def _autocovariancefn(x, max_lag, biased=False, axis=0):
 
     shape = list(y.shape)
     shape[-1] = max_lag
-    ac = zeros(shape, dtype=float_)
+    ac = zeros(shape, dtype=float64)
 
     for i in range(min(max_lag, y.shape[-1] - 10)):
         ac[..., i] = autocorrelation(y, i, True)
@@ -592,7 +592,7 @@ def __init__(self):
         )
 
     def _predict(self, fs, leg_length, gait, gait_aux):
-        pci = zeros(len(gait_aux["accel"]), dtype=float_)
+        pci = zeros(len(gait_aux["accel"]), dtype=float64)
 
         phase = gait["PARAM:step time"] / gait["PARAM:stride time"]  # %, not degrees
         for i in range(len(gait_aux["accel"])):
@@ -665,7 +665,7 @@ def __init__(self):
         super().__init__("gait symmetry index", __name__, depends=[StrideTime])
 
     def _predict(self, fs, leg_length, gait, gait_aux):
-        gsi = zeros(len(gait_aux["accel"]), dtype=float_)
+        gsi = zeros(len(gait_aux["accel"]), dtype=float64)
 
         # setup acceleration filter if its possible to use
         if 0 < (2 * 10 / fs) < 1:
@@ -741,7 +741,7 @@ def __init__(self):
         super().__init__("step regularity - V", __name__, depends=[StepTime])
 
     def _predict(self, fs, leg_length, gait, gait_aux):
-        stepreg = full(len(gait_aux["accel"]), nan, dtype=float_)
+        stepreg = full(len(gait_aux["accel"]), nan, dtype=float64)
 
         for i in unique(gait_aux["inertial data i"]):
             acc = gait_aux["accel"][i]
@@ -801,7 +801,7 @@ def __init__(self):
         super().__init__("stride regularity - V", __name__, depends=[StrideTime])
 
     def _predict(self, fs, leg_length, gait, gait_aux):
-        stridereg = full(len(gait_aux["accel"]), nan, dtype=float_)
+        stridereg = full(len(gait_aux["accel"]), nan, dtype=float64)
 
         for i in unique(gait_aux["inertial data i"]):
             acc = gait_aux["accel"][i]
diff --git a/src/skdh/io/empatica.py b/src/skdh/io/empatica.py
@@ -12,7 +12,7 @@
     argmin,
     abs,
     nan,
-    float_,
+    float64,
 )
 
 from skdh import BaseProcess, handle_process_returns
@@ -237,7 +237,7 @@ def handle_resampling(self, streams):
                 new_shape = list(stream["values"].shape)
                 new_shape[0] = acc_dict[self._acc].shape[0]
                 # create the full length data shape
-                rs_streams[name] = full(new_shape, nan, dtype=float_)
+                rs_streams[name] = full(new_shape, nan, dtype=float64)
                 # get the indices for the stream data
                 i1 = argmin(abs(acc_dict["time"] - stream["time"][0]))
                 i2 = i1 + stream["time"].size
diff --git a/src/skdh/preprocessing/calibrate.py b/src/skdh/preprocessing/calibrate.py
@@ -16,7 +16,7 @@
     abs,
     all as npall,
     around,
-    Inf,
+    inf,
     vstack,
     minimum,
     concatenate,
@@ -197,6 +197,7 @@ def predict(self, *, time, accel, fs=None, apply=True, temperature=None, **kwarg
             i_h += 1
 
         if apply and valid_calibration:
+            self.logger.info(f"Applying calibration: {offset=}, {scale=}, {temp_mean=}, {temp_scale=}")
             if temperature is None:
                 accel = (accel + offset) * scale
             else:
@@ -274,7 +275,7 @@ def _do_iterative_closest_point_fit(self, store):
         tmp_rm = (tmp_rm - tmp_mean).reshape((-1, 1))
 
         weights = ones(acc_rm.shape[0]) * 100
-        res = [Inf]
+        res = [inf]
         LR = LinearRegression()
 
         for niter in range(self.max_iter):
diff --git a/src/skdh/utility/activity_counts.py b/src/skdh/utility/activity_counts.py
@@ -5,7 +5,7 @@
 Copyright (c) 2022. Pfizer Inc. All rights reserved.
 """
 
-from numpy import array, repeat, abs, minimum, floor, float_
+from numpy import array, repeat, abs, minimum, floor, float64
 from scipy.signal import lfilter_zi, lfilter
 
 from skdh.utility.internal import apply_resample
@@ -26,7 +26,7 @@
         -0.006788163862310,
         0.000000000000000,
     ],
-    dtype=float_,
+    dtype=float64,
 )
 
 output_coef = array(
@@ -41,7 +41,7 @@
         0.01949130205890000000,
         0.00000000000000000000,
     ],
-    dtype=float_,
+    dtype=float64,
 )
 
 
diff --git a/src/skdh/utility/fragmentation_endpoints.py b/src/skdh/utility/fragmentation_endpoints.py
@@ -17,7 +17,7 @@
     sum,
     log,
     nan,
-    float_,
+    float64,
 )
 
 from skdh.utility.internal import rle
@@ -66,8 +66,8 @@ def gini(x, w=None, corr=True):
         sorted_x = x[sorted_indices]
         sorted_w = w[sorted_indices]
         # Force float dtype to avoid overflows
-        cumw = cumsum(sorted_w, dtype=float_)
-        cumxw = cumsum(sorted_x * sorted_w, dtype=float_)
+        cumw = cumsum(sorted_w, dtype=float64)
+        cumxw = cumsum(sorted_x * sorted_w, dtype=float64)
         g = sum(cumxw[1:] * cumw[:-1] - cumxw[:-1] * cumw[1:]) / (cumxw[-1] * cumw[-1])
         if corr:
             return g * x.size / (x.size - 1)
@@ -76,7 +76,7 @@ def gini(x, w=None, corr=True):
     else:
         sorted_x = sort(x)
         n = x.size
-        cumx = cumsum(sorted_x, dtype=float_)
+        cumx = cumsum(sorted_x, dtype=float64)
         # The above formula, with all weights equal to 1 simplifies to:
         g = (n + 1 - 2 * sum(cumx) / cumx[-1]) / n
         if corr:
diff --git a/src/skdh/utility/internal.py b/src/skdh/utility/internal.py
@@ -20,7 +20,7 @@
     around,
     diff,
     mean,
-    float_,
+    float64,
     int_,
     ndarray,
     concatenate,
@@ -218,7 +218,7 @@ def resample(x, factor, t, t_rs):
             if x.ndim == 1:
                 return (interp(t_rs, t, x),)
             elif x.ndim == 2:
-                xrs = zeros((t_rs.size, x.shape[1]), dtype=float_)
+                xrs = zeros((t_rs.size, x.shape[1]), dtype=float64)
                 for j in range(x.shape[1]):
                     xrs[:, j] = interp(t_rs, t, x[:, j])
                 return (xrs,)
diff --git a/src/skdh/utility/math.py b/src/skdh/utility/math.py
diff --git a/tests/sleep/test_sleep_classification.py b/tests/sleep/test_sleep_classification.py
diff --git a/tests/sleep/test_utilities.py b/tests/sleep/test_utilities.py

Original file line number	Diff line number	Diff line change
`@@ -1,7 +1,7 @@`
`1`	`1`	`project(`
`2`	`2`	`'scikit-digital-health',`
`3`	`3`	`'c',`
`4`		`- version: '0.16.2',`
	`4`	`+ version: '0.16.3',`
`5`	`5`	`license: 'MIT',`
`6`	`6`	`meson_version: '>=1.1',`
`7`	`7`	`)`