Decoder

class Decoder(categories: dict, n_splits: int = 5, n_repeats: int = 1, min_samples: int = 1, which: str = 'test', **kwargs)[source][source]
Parameters:

  • categories (dict)

  • n_splits (int)

  • n_repeats (int)

  • min_samples (int)

  • which (str)

cv_cm(x_data: Any, labels: Any, normalize: str = None, obs_axs: int = -2, n_jobs: int = 1, average_repetitions: bool = True, window: int = None, shuffle: bool = False, oversample: bool = True, step: int = 1) → Any[source][source]

Cross-validated confusion matrix

Parameters:

  • x_data (np.ndarray) – The data to be decoded

  • labels (np.ndarray) – The labels for the data

  • normalize (str, optional) – How to normalize the confusion matrix, by default None

  • obs_axs (int, optional) – The axis containing the observations, by default -2

  • n_jobs (int, optional) – The number of jobs to run in parallel, by default 1

  • average_repetitions (bool, optional) – Whether to average the repetitions, by default True

  • window (int, optional) – The window size for time sliding, by default None

  • shuffle (bool, optional) – Whether to shuffle the labels, by default False

  • oversample (bool, optional) – Whether to oversample the training data, by default True

  • step (int, optional) – The step size for time sliding, by default 1

Returns:

The confusion matrix

Return type:

np.ndarray

Examples

>>> np.random.seed(42)
>>> decoder = Decoder({'heat': 1, 'hoot': 2, 'hot': 3, 'hut': 4},
...             5, 10, explained_variance=0.8, da_type='lda')
>>> X = np.random.randn(100, 50, 100)
>>> labels = np.random.randint(1, 5, 50)
>>> decoder.cv_cm(X, labels, normalize='true')
array([[0.11111111, 0.        , 0.03333333, 0.85555556],
       [0.1       , 0.        , 0.04      , 0.86      ],
       [0.10666667, 0.        , 0.04      , 0.85333333],
       [0.10625   , 0.        , 0.03125   , 0.8625    ]])
>>> decoder = Decoder({'heat': 1, 'hoot': 2, 'hot': 3, 'hut': 4},
...             5, 10, explained_variance=0.8, da_type='lda')
>>> decoder.cv_cm(X, labels, normalize='true', window=20, step=5)[0]
array([[0.04444444, 0.        , 0.36666667, 0.58888889],
       [0.02      , 0.01      , 0.41      , 0.56      ],
       [0.03333333, 0.        , 0.50666667, 0.46      ],
       [0.03125   , 0.        , 0.5       , 0.46875   ]])
>>> decoder.cv_cm(X, labels, normalize='true', window=20, step=5,
...     shuffle=True, oversample=True)[0]
array([[0.        , 0.12222222, 0.52222222, 0.35555556],
       [0.01      , 0.12      , 0.5       , 0.37      ],
       [0.00666667, 0.10666667, 0.50666667, 0.38      ],
       [0.        , 0.09375   , 0.525     , 0.38125   ]])
>>> import cupy as cp
>>> X = cp.random.randn(100, 100, 50, 100)
>>> X[0, 0, 0, :] = np.nan
>>> labels = cp.random.randint(1, 5, 50)
>>> with config_context(array_api_dispatch=True):
...     decoder.cv_cm(X, labels, normalize='true')
array([[0.        , 0.36666667, 0.63333333, 0.        ],
       [0.        , 0.32777778, 0.67222222, 0.        ],
       [0.        , 0.33157895, 0.66842105, 0.        ],
       [0.        , 0.35714286, 0.64285714, 0.        ]])

Examples using ieeg.decoding.decode.Decoder

PCA-LDA Decoding

PCA-LDA Decoding