python-hiddenmarkov 0.3.0__cp310-cp310-win_amd64.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- hiddenmarkov/__init__.py +852 -0
- hiddenmarkov/_hmm.c +33227 -0
- hiddenmarkov/_hmm.cp310-win_amd64.pyd +0 -0
- hiddenmarkov/_hmm.pyi +10 -0
- hiddenmarkov/_hmm.pyx +90 -0
- hiddenmarkov/py.typed +0 -0
- hiddenmarkov/utils.py +55 -0
- python_hiddenmarkov-0.3.0.dist-info/METADATA +84 -0
- python_hiddenmarkov-0.3.0.dist-info/RECORD +12 -0
- python_hiddenmarkov-0.3.0.dist-info/WHEEL +5 -0
- python_hiddenmarkov-0.3.0.dist-info/licenses/LICENSE +21 -0
- python_hiddenmarkov-0.3.0.dist-info/top_level.txt +2 -0
|
Binary file
|
hiddenmarkov/_hmm.pyi
ADDED
hiddenmarkov/_hmm.pyx
ADDED
|
@@ -0,0 +1,90 @@
|
|
|
1
|
+
# cython: language_level=3
|
|
2
|
+
# cython: boundscheck=False
|
|
3
|
+
# cython: wraparound=False
|
|
4
|
+
# cython: cdivision=True
|
|
5
|
+
"""Cython-accelerated inner loop for the Viterbi algorithm.
|
|
6
|
+
|
|
7
|
+
Why Cython helps here but not in the forward step
|
|
8
|
+
--------------------------------------------------
|
|
9
|
+
The Viterbi step needs a row-wise argmax over the N×N matrix
|
|
10
|
+
omega[t-1, i] + log_T[i, j] for all (i, j).
|
|
11
|
+
NumPy handles this by broadcasting a length-N vector over the whole N×N
|
|
12
|
+
transition matrix, creating a fresh N×N temporary allocation at *every*
|
|
13
|
+
time step. The Cython version replaces that with a doubly-nested C loop:
|
|
14
|
+
no temporaries, working set stays in L1/L2 cache for moderate N.
|
|
15
|
+
|
|
16
|
+
The forward step is different: its inner operation is a plain matrix-vector
|
|
17
|
+
multiply (np.dot), which NumPy hands to BLAS. BLAS is already SIMD-
|
|
18
|
+
vectorised and for N ≳ 64 will outperform any pure-C scalar loop. We
|
|
19
|
+
therefore keep the forward step in NumPy.
|
|
20
|
+
"""
|
|
21
|
+
|
|
22
|
+
import numpy as np
|
|
23
|
+
cimport numpy as np
|
|
24
|
+
from libc.math cimport INFINITY
|
|
25
|
+
|
|
26
|
+
np.import_array()
|
|
27
|
+
|
|
28
|
+
|
|
29
|
+
def viterbi_loop(
|
|
30
|
+
double[:, ::1] omega,
|
|
31
|
+
np.int64_t[:, ::1] omega_idx,
|
|
32
|
+
const double[:, ::1] obs_probs,
|
|
33
|
+
const double[:, ::1] trans_mat,
|
|
34
|
+
const double[:] init_probs,
|
|
35
|
+
bint log_probabilities,
|
|
36
|
+
):
|
|
37
|
+
"""Fill the Viterbi trellis on pre-computed probability arrays.
|
|
38
|
+
|
|
39
|
+
The caller is responsible for pre-computing *obs_probs* and *trans_mat*
|
|
40
|
+
from the HMM's Python model objects, and for the backtracking step
|
|
41
|
+
after this function returns.
|
|
42
|
+
|
|
43
|
+
Parameters
|
|
44
|
+
----------
|
|
45
|
+
omega : (T, N) float64, written in-place
|
|
46
|
+
Viterbi trellis.
|
|
47
|
+
omega_idx : (T, N) int64, written in-place
|
|
48
|
+
Backtracking table; omega_idx[t, j] = best predecessor state for
|
|
49
|
+
state j at time t.
|
|
50
|
+
obs_probs : (T, N) float64
|
|
51
|
+
Observation (log-)probabilities for every time step and state.
|
|
52
|
+
trans_mat : (N, N) float64
|
|
53
|
+
Transition (log-)probabilities; trans_mat[i, j] = P(j | i).
|
|
54
|
+
init_probs : (N,) float64
|
|
55
|
+
Initial state (log-)probabilities.
|
|
56
|
+
log_probabilities : bool
|
|
57
|
+
When True all arrays contain log-probabilities and additions replace
|
|
58
|
+
multiplications; when False they contain raw probabilities.
|
|
59
|
+
"""
|
|
60
|
+
cdef Py_ssize_t T = obs_probs.shape[0]
|
|
61
|
+
cdef Py_ssize_t N = obs_probs.shape[1]
|
|
62
|
+
cdef Py_ssize_t t, i, j
|
|
63
|
+
cdef double best_val, val
|
|
64
|
+
|
|
65
|
+
if log_probabilities:
|
|
66
|
+
for j in range(N):
|
|
67
|
+
omega[0, j] = obs_probs[0, j] + init_probs[j]
|
|
68
|
+
for t in range(1, T):
|
|
69
|
+
for j in range(N):
|
|
70
|
+
best_val = -INFINITY
|
|
71
|
+
omega_idx[t, j] = 0
|
|
72
|
+
for i in range(N):
|
|
73
|
+
val = omega[t - 1, i] + trans_mat[i, j]
|
|
74
|
+
if val > best_val:
|
|
75
|
+
best_val = val
|
|
76
|
+
omega_idx[t, j] = i
|
|
77
|
+
omega[t, j] = obs_probs[t, j] + best_val
|
|
78
|
+
else:
|
|
79
|
+
for j in range(N):
|
|
80
|
+
omega[0, j] = obs_probs[0, j] * init_probs[j]
|
|
81
|
+
for t in range(1, T):
|
|
82
|
+
for j in range(N):
|
|
83
|
+
best_val = 0.0
|
|
84
|
+
omega_idx[t, j] = 0
|
|
85
|
+
for i in range(N):
|
|
86
|
+
val = omega[t - 1, i] * trans_mat[i, j]
|
|
87
|
+
if val > best_val:
|
|
88
|
+
best_val = val
|
|
89
|
+
omega_idx[t, j] = i
|
|
90
|
+
omega[t, j] = obs_probs[t, j] * best_val
|
hiddenmarkov/py.typed
ADDED
|
File without changes
|
hiddenmarkov/utils.py
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
1
|
+
#!/usr/bin/env python
|
|
2
|
+
# -*- coding: utf-8 -*-
|
|
3
|
+
"""
|
|
4
|
+
Utility functions
|
|
5
|
+
"""
|
|
6
|
+
from __future__ import annotations
|
|
7
|
+
|
|
8
|
+
from typing import Any, Callable, List, Tuple
|
|
9
|
+
|
|
10
|
+
import numpy as np
|
|
11
|
+
|
|
12
|
+
|
|
13
|
+
def inverted_softmax(array: np.ndarray) -> np.ndarray:
|
|
14
|
+
"""Convert distances to probabilities: lower distance → higher probability."""
|
|
15
|
+
e = np.exp(-array)
|
|
16
|
+
return e / e.sum()
|
|
17
|
+
|
|
18
|
+
|
|
19
|
+
def softmax(array: np.ndarray) -> np.ndarray:
|
|
20
|
+
"""Convert scores to probabilities: higher score → higher probability."""
|
|
21
|
+
e = np.exp(array)
|
|
22
|
+
return e / e.sum()
|
|
23
|
+
|
|
24
|
+
|
|
25
|
+
def create_prob_models(
|
|
26
|
+
no_global_states: int = 100,
|
|
27
|
+
window_size: int = 5,
|
|
28
|
+
) -> List[List[Tuple[Callable[..., Any], int]]]:
|
|
29
|
+
"""
|
|
30
|
+
Build example windowed probability models for use with
|
|
31
|
+
:class:`~hiddenmarkov.WindowedObservationModel` and
|
|
32
|
+
:func:`~hiddenmarkov.viterbi_algorithm_windowed`.
|
|
33
|
+
|
|
34
|
+
Parameters
|
|
35
|
+
----------
|
|
36
|
+
no_global_states : int
|
|
37
|
+
Total number of global states in the HMM.
|
|
38
|
+
window_size : int
|
|
39
|
+
Number of states visible at any one time.
|
|
40
|
+
|
|
41
|
+
Returns
|
|
42
|
+
-------
|
|
43
|
+
list of lists of (callable, int) pairs
|
|
44
|
+
``result[i]`` is the window of ``(prob_model, global_index)`` pairs
|
|
45
|
+
centred at global state ``i``.
|
|
46
|
+
"""
|
|
47
|
+
models_per_state: List[List[Tuple[Callable[..., Any], int]]] = []
|
|
48
|
+
for i in range(no_global_states):
|
|
49
|
+
models: List[Tuple[Callable[..., Any], int]] = []
|
|
50
|
+
for j in range(window_size):
|
|
51
|
+
def probf(obs: Any, ref: int = i + j) -> int:
|
|
52
|
+
return int(obs == ref)
|
|
53
|
+
models.append((probf, max(0, min(no_global_states - 1, i + j))))
|
|
54
|
+
models_per_state.append(models)
|
|
55
|
+
return models_per_state
|
|
@@ -0,0 +1,84 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: python-hiddenmarkov
|
|
3
|
+
Version: 0.3.0
|
|
4
|
+
Summary: A simple Python package for Hidden Markov Models
|
|
5
|
+
Author: Silvan Peter
|
|
6
|
+
Author-email: Carlos Cancino-Chacón <carloscancinochacon@gmail.com>
|
|
7
|
+
License: MIT
|
|
8
|
+
Project-URL: Homepage, https://github.com/CarlosCancino-Chacon/hiddenmarkov
|
|
9
|
+
Keywords: hmm,hidden_markov_models
|
|
10
|
+
Classifier: License :: OSI Approved :: MIT License
|
|
11
|
+
Classifier: Programming Language :: Python
|
|
12
|
+
Classifier: Programming Language :: Python :: 3
|
|
13
|
+
Classifier: Programming Language :: Python :: Implementation :: CPython
|
|
14
|
+
Classifier: Programming Language :: Python :: Implementation :: PyPy
|
|
15
|
+
Requires-Python: >=3.10
|
|
16
|
+
Description-Content-Type: text/markdown
|
|
17
|
+
License-File: LICENSE
|
|
18
|
+
Requires-Dist: numpy>=1.24
|
|
19
|
+
Requires-Dist: scipy>=1.7
|
|
20
|
+
Dynamic: license-file
|
|
21
|
+
|
|
22
|
+
# hiddenmarkov
|
|
23
|
+
|
|
24
|
+
[](https://pypi.org/project/python-hiddenmarkov/)
|
|
25
|
+
[](https://github.com/CarlosCancino-Chacon/hiddenmarkov/actions/workflows/ci.yml)
|
|
26
|
+
[](https://codecov.io/gh/CarlosCancino-Chacon/hiddenmarkov)
|
|
27
|
+
|
|
28
|
+
A simple Python library for Hidden Markov Models.
|
|
29
|
+
|
|
30
|
+
This library was created for educational purposes and does not aim to have the most efficient implementation.
|
|
31
|
+
|
|
32
|
+
## Setup
|
|
33
|
+
|
|
34
|
+
Simple installation
|
|
35
|
+
|
|
36
|
+
```
|
|
37
|
+
pip install python-hiddenmarkov
|
|
38
|
+
```
|
|
39
|
+
|
|
40
|
+
Develop version
|
|
41
|
+
```bash
|
|
42
|
+
git clone https://github.com/CarlosCancino-Chacon/hiddenmarkov.git
|
|
43
|
+
cd hiddenmarkov
|
|
44
|
+
pip install -e .
|
|
45
|
+
```
|
|
46
|
+
|
|
47
|
+
## Usage
|
|
48
|
+
|
|
49
|
+
Example from [Wikipedia](https://en.wikipedia.org/wiki/Viterbi_algorithm#Example)
|
|
50
|
+
|
|
51
|
+
```python
|
|
52
|
+
import numpy as np
|
|
53
|
+
from hiddenmarkov import CategoricalObservationModel, ConstantTransitionModel, HMM
|
|
54
|
+
|
|
55
|
+
obs = ("normal", "cold", "dizzy")
|
|
56
|
+
observations = ("normal", "cold", "dizzy")
|
|
57
|
+
states = ("Healthy", "Fever")
|
|
58
|
+
observation_probabilities = np.array([[0.5, 0.1], [0.4, 0.3], [0.1, 0.6]])
|
|
59
|
+
transition_probabilities = np.array([[0.7, 0.3], [0.4, 0.6]])
|
|
60
|
+
|
|
61
|
+
observation_model = CategoricalObservationModel(
|
|
62
|
+
observation_probabilities, obs
|
|
63
|
+
)
|
|
64
|
+
|
|
65
|
+
init_distribution = np.array([0.6, 0.4])
|
|
66
|
+
|
|
67
|
+
transition_model = ConstantTransitionModel(
|
|
68
|
+
transition_probabilities, init_distribution
|
|
69
|
+
)
|
|
70
|
+
|
|
71
|
+
hmm = HMM(observation_model, transition_model, state_space=states)
|
|
72
|
+
|
|
73
|
+
path, prob = hmm.find_best_sequence(observations, log_probabilities=False)
|
|
74
|
+
print("Example Wikipedia")
|
|
75
|
+
print("Best sequence", path)
|
|
76
|
+
print("Expected Sequence", ["Healthy", "Healthy", "Fever"])
|
|
77
|
+
print("Sequence probability", prob)
|
|
78
|
+
print("Expected probability", 0.01512)
|
|
79
|
+
```
|
|
80
|
+
|
|
81
|
+
## Licence
|
|
82
|
+
|
|
83
|
+
The code in this package is licensed under the MIT Licence. For details,
|
|
84
|
+
please see the [LICENSE](https://github.com/CarlosCancino-Chacon/hiddenmarkov/blob/main/LICENSE) file.
|
|
@@ -0,0 +1,12 @@
|
|
|
1
|
+
hiddenmarkov/__init__.py,sha256=bqUzorpLu3mqL4v_Hv85VPLtme2qyNdOFvkzYYUD0Jo,31952
|
|
2
|
+
hiddenmarkov/_hmm.c,sha256=yVQeBYbTg_dY1978zwZxRIFuHAyYM3y65Uiv62ZuOCI,1239935
|
|
3
|
+
hiddenmarkov/_hmm.cp310-win_amd64.pyd,sha256=Wr6Pu3xbtUSrUuWoBnd6894v1D66Z6NGyHrk1OO93YA,143872
|
|
4
|
+
hiddenmarkov/_hmm.pyi,sha256=9AZfiDMoLRemrjIJ0jShItanFYb0jCJFz3LHYI0CNoI,224
|
|
5
|
+
hiddenmarkov/_hmm.pyx,sha256=rFwia_YlCzOGU2Ju6dlPTQo5165b9lrf9J-BxBSDA2U,3440
|
|
6
|
+
hiddenmarkov/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
|
|
7
|
+
hiddenmarkov/utils.py,sha256=t0dcXzlkxmEYGyxsrCDJzIq8WnfoAOumJo0n-LF9hQc,1717
|
|
8
|
+
python_hiddenmarkov-0.3.0.dist-info/licenses/LICENSE,sha256=aKW387ClCasMelXOka-SlIQnfJkO0sHLZ3LZABEfM8s,1122
|
|
9
|
+
python_hiddenmarkov-0.3.0.dist-info/METADATA,sha256=2dCAGU2aCOO9WowZWAtViimH9Qy85RDm_YS6nqUIO2U,2879
|
|
10
|
+
python_hiddenmarkov-0.3.0.dist-info/WHEEL,sha256=4_46qF9OLPzM_4q33Cq00hQdDDvlGpNmdPeJhAlTjEk,101
|
|
11
|
+
python_hiddenmarkov-0.3.0.dist-info/top_level.txt,sha256=1LNxieTtshGSqg1-YsRq7yLTvjWtKokyThxOhnDhKBs,24
|
|
12
|
+
python_hiddenmarkov-0.3.0.dist-info/RECORD,,
|
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
MIT License
|
|
2
|
+
|
|
3
|
+
Copyright (c) 2022 Carlos Eduardo Cancino-Chacón, Silvan Peter
|
|
4
|
+
|
|
5
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy
|
|
6
|
+
of this software and associated documentation files (the "Software"), to deal
|
|
7
|
+
in the Software without restriction, including without limitation the rights
|
|
8
|
+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
|
9
|
+
copies of the Software, and to permit persons to whom the Software is
|
|
10
|
+
furnished to do so, subject to the following conditions:
|
|
11
|
+
|
|
12
|
+
The above copyright notice and this permission notice shall be included in all
|
|
13
|
+
copies or substantial portions of the Software.
|
|
14
|
+
|
|
15
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
16
|
+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
17
|
+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
|
18
|
+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
19
|
+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
|
20
|
+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
|
21
|
+
SOFTWARE.
|