rapidtide 3.1__py3-none-any.whl → 3.1.1__py3-none-any.whl

rapidtide/dlfilter.py

@@ -26,6 +26,7 @@ import warnings
 import matplotlib as mpl
 import matplotlib.pyplot as plt
 import numpy as np
+from numpy.typing import NDArray
 
 with warnings.catch_warnings():
     warnings.simplefilter("ignore")
@@ -495,20 +496,20 @@ class DeepLearningFilter:
             )
 
     @tf.function
-    def predict_model(self, X: np.ndarray) -> np.ndarray:
+    def predict_model(self, X: NDArray) -> NDArray:
         """
         Make predictions using the trained model.
 
         Parameters
         ----------
-        X : np.ndarray
+        X : NDArray
             Input features for prediction. Shape should be (n_samples, n_features)
             where n_samples is the number of samples and n_features is the number
             of features expected by the model.
 
         Returns
         -------
-        np.ndarray
+        NDArray
             Model predictions. Shape will depend on the specific model type but
             typically follows (n_samples,) for regression or (n_samples, n_classes)
             for classification.
@@ -964,7 +965,7 @@ class DeepLearningFilter:
         self.savemodel()
         self.trained = True
 
-    def apply(self, inputdata: np.ndarray, badpts: np.ndarray | None = None) -> np.ndarray:
+    def apply(self, inputdata: NDArray, badpts: NDArray | None = None) -> NDArray:
         """
         Apply a sliding-window prediction model to the input data, optionally incorporating bad points.
 
@@ -975,15 +976,15 @@ class DeepLearningFilter:
 
         Parameters
         ----------
-        inputdata : np.ndarray
+        inputdata : NDArray
             Input data array of shape (N,) to be processed.
-        badpts : np.ndarray | None, optional
+        badpts : NDArray | None, optional
             Array of same shape as `inputdata` indicating bad or invalid points. If None, no bad points
             are considered. Default is None.
 
         Returns
         -------
-        np.ndarray
+        NDArray
             Predicted data array of the same shape as `inputdata`, with predictions aggregated and
             weighted across overlapping windows.
 
@@ -2813,14 +2814,14 @@ class HybridDLFilter(DeepLearningFilter):
 
 
 def filtscale(
-    data: np.ndarray,
+    data: NDArray,
     scalefac: float = 1.0,
     reverse: bool = False,
     hybrid: bool = False,
     lognormalize: bool = True,
     epsilon: float = 1e-10,
     numorders: int = 6,
-) -> tuple[np.ndarray, float] | np.ndarray:
+) -> tuple[NDArray, float] | NDArray:
     """
     Apply or reverse a frequency-domain scaling and normalization to input data.
 
@@ -2831,7 +2832,7 @@ def filtscale(
 
     Parameters
     ----------
-    data : np.ndarray
+    data : NDArray
         Input signal or transformed data (depending on `reverse` flag).
     scalefac : float, optional
         Scaling factor used for normalization. Default is 1.0.
@@ -2851,7 +2852,7 @@ def filtscale(
 
     Returns
     -------
-    tuple[np.ndarray, float] or np.ndarray
+    tuple[NDArray, float] or NDArray
         - If `reverse=False`: A tuple of (transformed data, scale factor).
           The transformed data is a stacked array of magnitude and phase components
           (or original data in hybrid mode).
@@ -3025,7 +3026,7 @@ def getmatchedtcs(
     >>> matched_files, length = getmatchedtcs("sub-*/func/*cardiacfromfmri_25.0Hz*")
     >>> print(f"Found {len(matched_files)} files with {length} timepoints each.")
     """
-    # list all of the target files
+    # list all the target files
     fromfile = sorted(glob.glob(searchstring))
     if debug:
         print(f"searchstring: {searchstring} -> {fromfile}")
@@ -3073,9 +3074,7 @@ def readindata(
     readlim: int | None = None,
     readskip: int | None = None,
     debug: bool = False,
-) -> (
-    tuple[np.ndarray, np.ndarray, list[str]] | tuple[np.ndarray, np.ndarray, list[str], np.ndarray]
-):
+) -> tuple[NDArray, NDArray, list[str]] | tuple[NDArray, NDArray, list[str], NDArray]:
     """
     Read and process time-series data from a list of matched files.
 
@@ -3112,7 +3111,7 @@ def readindata(
 
     Returns
     -------
-    tuple of (np.ndarray, np.ndarray, list[str]) or (np.ndarray, np.ndarray, list[str], np.ndarray)
+    tuple of (NDArray, NDArray, list[str]) or (NDArray, NDArray, list[str], NDArray)
         - `x1[startskip:-endskip, :count]`: Array of x-axis time series data.
         - `y1[startskip:-endskip, :count]`: Array of y-axis time series data.
         - `names[:count]`: List of file names that passed quality checks.
@@ -3174,7 +3173,7 @@ def readindata(
 
         Returns
         -------
-        tuple of (np.ndarray, np.ndarray, list[str]) or (np.ndarray, np.ndarray, list[str], np.ndarray)
+        tuple of (NDArray, NDArray, list[str]) or (NDArray, NDArray, list[str], NDArray)
             - `x1[startskip:-endskip, :count]`: Array of x-axis time series data.
             - `y1[startskip:-endskip, :count]`: Array of y-axis time series data.
             - `names[:count]`: List of file names that passed quality checks.
@@ -3366,8 +3365,8 @@ def prep(
     countlim: int | None = None,
     debug: bool = False,
 ) -> (
-    tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int]
-    | tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int, np.ndarray, np.ndarray]
+    tuple[NDArray, NDArray, NDArray, NDArray, int, int, int]
+    | tuple[NDArray, NDArray, NDArray, NDArray, int, int, int, NDArray, NDArray]
 ):
     """
     Prepare time-series data for training and validation by reading, normalizing,
@@ -3422,7 +3421,7 @@ def prep(
 
     Returns
     -------
-    tuple of (np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int)
+    tuple of (NDArray, NDArray, NDArray, NDArray, int, int, int)
         If `dofft` is False:
             - train_x : ndarray of shape (n_train, window_size, 1)
             - train_y : ndarray of shape (n_train, window_size, 1)
@@ -3432,8 +3431,8 @@ def prep(
             - tclen : int
             - batchsize : int
 
-        tuple of (np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int,
-                  np.ndarray, np.ndarray)
+        tuple of (NDArray, NDArray, NDArray, NDArray, int, int, int,
+                  NDArray, NDArray)
         If `dofft` is True:
             - train_x : ndarray of shape (n_train, window_size, 2)
             - train_y : ndarray of shape (n_train, window_size, 2)

rapidtide/dlfiltertorch.py

@@ -26,6 +26,7 @@ import matplotlib as mpl
 import matplotlib.pyplot as plt
 import numpy as np
 import tqdm
+from numpy.typing import NDArray
 
 with warnings.catch_warnings():
     warnings.simplefilter("ignore")
@@ -395,20 +396,20 @@ class DeepLearningFilter:
                 countlim=self.countlim,
             )
 
-    def predict_model(self, X: np.ndarray) -> np.ndarray:
+    def predict_model(self, X: NDArray) -> NDArray:
         """
         Make predictions using the trained model.
 
         Parameters
         ----------
-        X : np.ndarray
+        X : NDArray
             Input features for prediction. Shape should be (n_samples, n_features)
             where n_samples is the number of samples and n_features is the number
             of features expected by the model.
 
         Returns
         -------
-        np.ndarray
+        NDArray
             Model predictions. Shape will depend on the specific model type but
             typically follows (n_samples,) for regression or (n_samples, n_classes)
             for classification.
@@ -1067,7 +1068,7 @@ class DeepLearningFilter:
         self.savemodel()
         self.trained = True
 
-    def apply(self, inputdata: np.ndarray, badpts: np.ndarray | None = None) -> np.ndarray:
+    def apply(self, inputdata: NDArray, badpts: NDArray | None = None) -> NDArray:
         """
         Apply a sliding-window prediction model to the input data, optionally incorporating bad points.
 
@@ -1078,15 +1079,15 @@ class DeepLearningFilter:
 
         Parameters
         ----------
-        inputdata : np.ndarray
+        inputdata : NDArray
             Input data array of shape (N,) to be processed.
-        badpts : np.ndarray | None, optional
+        badpts : NDArray | None, optional
             Array of same shape as `inputdata` indicating bad or invalid points. If None, no bad points
             are considered. Default is None.
 
         Returns
         -------
-        np.ndarray
+        NDArray
             Predicted data array of the same shape as `inputdata`, with predictions aggregated and
             weighted across overlapping windows.
 
@@ -3958,14 +3959,14 @@ class HybridDLFilter(DeepLearningFilter):
 
 
 def filtscale(
-    data: np.ndarray,
+    data: NDArray,
     scalefac: float = 1.0,
     reverse: bool = False,
     hybrid: bool = False,
     lognormalize: bool = True,
     epsilon: float = 1e-10,
     numorders: int = 6,
-) -> tuple[np.ndarray, float] | np.ndarray:
+) -> tuple[NDArray, float] | NDArray:
     """
     Apply or reverse a scaling transformation to spectral data.
 
@@ -3977,7 +3978,7 @@ def filtscale(
 
     Parameters
     ----------
-    data : np.ndarray
+    data : NDArray
         Input time-domain signal or scaled spectral data depending on `reverse` flag.
     scalefac : float, optional
         Scaling factor used in normalization. Default is 1.0.
@@ -3996,7 +3997,7 @@ def filtscale(
 
     Returns
     -------
-    tuple[np.ndarray, float] or np.ndarray
+    tuple[NDArray, float] or NDArray
         - If `reverse` is False: Returns a tuple of (scaled_data, scalefac).
           `scaled_data` is a stacked array of magnitude and phase (or original signal
           and magnitude in hybrid mode).
@@ -4217,9 +4218,7 @@ def readindata(
     readlim: int | None = None,
     readskip: int | None = None,
     debug: bool = False,
-) -> (
-    tuple[np.ndarray, np.ndarray, list[str]] | tuple[np.ndarray, np.ndarray, list[str], np.ndarray]
-):
+) -> tuple[NDArray, NDArray, list[str]] | tuple[NDArray, NDArray, list[str], NDArray]:
     """
     Read and process time-series data from a list of matched files.
 
@@ -4258,7 +4257,7 @@ def readindata(
 
     Returns
     -------
-    tuple of (np.ndarray, np.ndarray, list[str]) or (np.ndarray, np.ndarray, list[str], np.ndarray)
+    tuple of (NDArray, NDArray, list[str]) or (NDArray, NDArray, list[str], NDArray)
         - `x1`: Array of shape `(tclen, count)` containing x-time series data.
         - `y1`: Array of shape `(tclen, count)` containing y-time series data.
         - `names`: List of file names that passed quality checks.
@@ -4441,8 +4440,8 @@ def prep(
     countlim: int | None = None,
     debug: bool = False,
 ) -> (
-    tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int]
-    | tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int, np.ndarray, np.ndarray]
+    tuple[NDArray, NDArray, NDArray, NDArray, int, int, int]
+    | tuple[NDArray, NDArray, NDArray, NDArray, int, int, int, NDArray, NDArray]
 ):
     """
     Prepare time-series data for training and validation by reading, normalizing,
@@ -4497,7 +4496,7 @@ def prep(
 
     Returns
     -------
-    tuple of (np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int)
+    tuple of (NDArray, NDArray, NDArray, NDArray, int, int, int)
         If `dofft` is False:
             - train_x : Training input data (shape: [n_windows, window_size, 1])
             - train_y : Training target data (shape: [n_windows, window_size, 1])
@@ -4507,7 +4506,7 @@ def prep(
             - tclen : Total time points after skipping
             - batchsize : Number of windows per subject
 
-        tuple of (np.ndarray, np.ndarray, np.ndarray, np.ndarray, int, int, int, np.ndarray, np.ndarray)
+        tuple of (NDArray, NDArray, NDArray, NDArray, int, int, int, NDArray, NDArray)
         If `dofft` is True:
             - train_x : Training input data (shape: [n_windows, window_size, 2])
             - train_y : Training target data (shape: [n_windows, window_size, 2])

rapidtide/filter.py

@@ -1513,9 +1513,9 @@ def wiener_deconvolution(signal: NDArray, kernel: NDArray, lambd: float) -> NDAr
 
     Parameters
     ----------
-    signal : numpy.NDArray
+    signal : NDArray
         Input signal to be deconvolved, 1D array.
-    kernel : numpy.NDArray
+    kernel : NDArray
         Convolution kernel (point spread function), 1D array.
     lambd : float
         Regularization parameter representing the signal-to-noise ratio in
@@ -1523,7 +1523,7 @@ def wiener_deconvolution(signal: NDArray, kernel: NDArray, lambd: float) -> NDAr
 
     Returns
     -------
-    numpy.NDArray
+    NDArray
         Deconvolved signal, same length as input signal.
 
     Notes
@@ -1601,7 +1601,7 @@ def spectrum(
 
     Parameters
     ----------
-    spectrum : numpy.NDArray
+    spectrum : NDArray
         Input spectrum array. Should contain non-negative values.
 
     Returns

rapidtide/fit.py

@@ -1056,13 +1056,13 @@ def territorydecomp(
     ----------
     inputmap : NDArray[np.floating[Any]]
         Input data to be decomposed. Can be 3D or 4D (e.g., time series).
-    template : numpy.ndarray
+    template : NDArray
         Template values corresponding to the spatial locations in `inputmap`.
         Should have the same shape as `inputmap` (or be broadcastable).
-    atlas : numpy.ndarray
+    atlas : NDArray
         Atlas defining the territories. Each unique integer value represents a distinct region.
         Must have the same shape as `inputmap`.
-    inputmask : numpy.ndarray, optional
+    inputmask : NDArray, optional
         Mask to define valid voxels in `inputmap`. If None, all voxels are considered valid.
         Should have the same shape as `inputmap`.
     intercept : bool, optional
@@ -1074,14 +1074,14 @@ def territorydecomp(
 
     Returns
     -------
-    tuple of numpy.ndarray
+    tuple of NDArray
         A tuple containing:
-        - fitmap : numpy.ndarray
+        - fitmap : NDArray
             The decomposed map with fitted values projected back onto the original spatial locations.
-        - thecoffs : numpy.ndarray
+        - thecoffs : NDArray
             Array of polynomial coefficients for each territory and map. Shape is (nummaps, numterritories, fitorder+1)
             if `intercept` is True, or (nummaps, numterritories, fitorder) otherwise.
-        - theR2s : numpy.ndarray
+        - theR2s : NDArray
             R-squared values for the fits for each territory and map. Shape is (nummaps, numterritories).
 
     Notes
@@ -2232,7 +2232,7 @@ def gram_schmidt(theregressors: NDArray, debug: bool = False) -> NDArray:
 
     Parameters
     ----------
-    theregressors : numpy.ndarray
+    theregressors : NDArray
         A 2D NumPy array where each row represents a vector to be orthogonalized.
     debug : bool, optional
         If True, prints debug information about input and output dimensions.
@@ -2240,7 +2240,7 @@ def gram_schmidt(theregressors: NDArray, debug: bool = False) -> NDArray:
 
     Returns
     -------
-    numpy.ndarray
+    NDArray
         A 2D NumPy array representing the orthonormal basis. Each row is an
         orthonormal vector. The number of rows may be less than the input if
         some vectors were linearly dependent.
@@ -2292,7 +2292,7 @@ def mlproject(thefit: NDArray, theevs: list, intercept: bool) -> NDArray:
         A 1D array or list of coefficients (weights) to be applied to the
         explanatory variables. If `intercept` is True, the first element of
         `thefit` is treated as the intercept.
-    theevs : list of numpy.ndarray
+    theevs : list of NDArray
         A list where each element is a 1D NumPy array representing an
         explanatory variable (feature time series). The length of `theevs`
         should match the number of non-intercept coefficients in `thefit`.
@@ -2510,9 +2510,9 @@ def calcexpandedregressors(
 
     Returns
     -------
-    tuple of (numpy.ndarray, list)
+    tuple of (NDArray, list)
         A tuple containing:
-        - outputregressors : numpy.ndarray
+        - outputregressors : NDArray
           A 2D NumPy array where each row represents a generated regressor
           (original, polynomial, or derivative) and columns represent time points.
         - outlabels : list of str
@@ -2587,6 +2587,7 @@ def calcexpandedregressors(
             activecolumn += 1
     return outputregressors, outlabels
 
+
 @conditionaljit()
 def derivativelinfitfilt(
     thedata: NDArray, theevs: NDArray, nderivs: int = 1, debug: bool = False
@@ -2670,6 +2671,7 @@ def derivativelinfitfilt(
 
     return filtered, thenewevs, datatoremove, R, coffs
 
+
 @conditionaljit()
 def expandedlinfitfilt(
     thedata: NDArray, theevs: NDArray, ncomps: int = 1, debug: bool = False
@@ -2752,6 +2754,7 @@ def expandedlinfitfilt(
 
     return filtered, thenewevs, datatoremove, R, coffs
 
+
 @conditionaljit()
 def linfitfilt(
     thedata: NDArray, theevs: NDArray, debug: bool = False
@@ -2843,8 +2846,7 @@ def confoundregress(
     regressors: NDArray,
     debug: bool = False,
     showprogressbar: bool = True,
-    rt_floatset: type = np.float64,
-    rt_floattype: str = "float64",
+    rt_floattype: np.dtype = np.float64,
 ) -> Tuple[NDArray, NDArray]:
     """
     Filters multiple regressors out of an array of data using linear regression.
@@ -2864,10 +2866,8 @@ def confoundregress(
         Print additional diagnostic information if True. Default is False.
     showprogressbar : bool, optional
         Show progress bar during processing. Default is True.
-    rt_floatset : type, optional
+    rt_floattype : np.dtype, optional
         The data type used for floating-point calculations. Default is np.float64.
-    rt_floattype : str, optional
-        The string representation of the floating-point data type. Default is "float64".
 
     Returns
     -------
@@ -2905,7 +2905,7 @@ def confoundregress(
         if i == 0 and debug:
             print("fit shape:", thefit.shape)
         for j in range(regressors.shape[0]):
-            datatoremove += rt_floatset(rt_floatset(thefit[0, 1 + j]) * regressors[j, :])
+            datatoremove += (thefit[0, 1 + j] * regressors[j, :]).astype(rt_floattype)
         filtereddata[i, :] = data[i, :] - datatoremove
         r2value[i] = R2
     return filtereddata, r2value