-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Kaleb
committed
Apr 7, 2020
1 parent
f25c5d5
commit 237ce91
Showing
13 changed files
with
441 additions
and
19 deletions.
There are no files selected for viewing
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Binary file not shown.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,174 @@ | ||
| import numpy as np | ||
| import copy | ||
| import pickle | ||
| import os | ||
| import copy | ||
|
|
||
| from sklearn.linear_model import LogisticRegression | ||
|
|
||
| from speech_module import SpeechModule | ||
| from base_classes import Object, Context | ||
|
|
||
| class SpeechLearner: | ||
| def __init__(self): | ||
| self.clf = LogisticRegression(max_iter=1000) | ||
|
|
||
| def train(self, X, y): | ||
| self.clf.fit(X, y) | ||
| print(self.clf.classes_) | ||
|
|
||
| def predict(self, X): | ||
| return self.clf.predict(X) | ||
|
|
||
| def predict_probs(self, X): | ||
| return self.clf.predict_proba(X) | ||
|
|
||
| def plot_learned_function(self, data): | ||
| # TODO implement | ||
| raise NotImplementedError | ||
| pass | ||
|
|
||
| def save_model(self, filename): | ||
| pickle.dump(self.clf, open(filename, 'wb')) | ||
|
|
||
| def load_model(self, filename): | ||
| self.clf = pickle.load(open(filename, 'rb')) | ||
|
|
||
| class REG: | ||
| def __init__(self): | ||
| w2c = "data/w2c_4096.txt" | ||
| # w2c = "/ros/catkin_ws/src/hrc_discrim_learning/src/hrc_discrim_learning/data/w2c_4096.txt" | ||
| self.sm = SpeechModule(w2c) | ||
| self.theta = 0.5 | ||
|
|
||
| self.features = ["color", "size", "dim", "none"] | ||
|
|
||
| self.model = SpeechLearner() | ||
|
|
||
| def save_models(self, filename="model/lgreg.pkl"): | ||
| self.model.save_model(filename) | ||
|
|
||
| def load_models(self, filename="model/lgreg.pkl"): | ||
| self.model.load_model(filename) | ||
| self.mapping = self.model.clf.classes_ | ||
|
|
||
| def train_model(self, x, y): | ||
| self.model.train(x, y) | ||
| self.mapping = self.model.clf.classes_ | ||
|
|
||
| def _generate_single_output(self, object, context, type_env, feature_set): | ||
| """ UPDATED """ | ||
| features= copy.copy(feature_set) | ||
| model_input, labels, results = self.get_model_input(object, context) | ||
| # prediction = self.mapping[self.model.predict([model_input])] | ||
| pdist = self.model.predict_probs([model_input])[0] | ||
| # map prob distributions as a dict | ||
| pdist_as_dict = {} | ||
| for i in range(4): | ||
| cls = self.mapping[i] | ||
| pdist_as_dict[cls] = pdist[i] | ||
|
|
||
| if len(type_env) > 1 and "none" : | ||
| features.remove("none") | ||
| best = max(feature_set, key=lambda x: pdist_as_dict[x]) | ||
|
|
||
| if best == "none": | ||
| return None, None, None | ||
| else: | ||
| return best, labels[best], results[best] | ||
|
|
||
| def generate_output(self, object, context): | ||
| # context should include object | ||
| """ UPDATED """ | ||
|
|
||
| output = "" | ||
| type = object.get_feature_val('type') | ||
| type_env = context.get_type_match(type) | ||
|
|
||
| feature_set = copy.copy(self.features) | ||
| while feature_set: | ||
| feature, label, new_context = self._generate_single_output(object, context, type_env, feature_set) | ||
| if not label: | ||
| break | ||
| output += (label + " ") | ||
| context = Context(new_context) | ||
| type_env = context.get_type_match(type) | ||
| feature_set.remove(feature) | ||
|
|
||
| output+=type | ||
| return output | ||
|
|
||
|
|
||
| # output = "" | ||
| # | ||
| # type = object.get_feature_val("type") | ||
| # type_env = context.get_type_match(type) | ||
| # | ||
| # # next: iterate through possible features | ||
| # feature_set = copy.copy(self.features) | ||
| # while feature_set: | ||
| # feature, label, new_context = self._generate_single_output(object, context, type_env, feature_set) | ||
| # if not label: | ||
| # output += type | ||
| # return output | ||
| # | ||
| # output += (label + " ") | ||
| # context = Context(new_context) | ||
| # type_env = context.get_type_match(type) | ||
| # feature_set.remove(feature) | ||
| # | ||
| # # return "ERR: check REGenerator" | ||
| # return output | ||
|
|
||
| def get_model_input(self, object, context): | ||
| v = [] | ||
| labels = {} | ||
| results = {} | ||
| for feature in self.features[:-1]: | ||
| label, score, data, kept_objects = self._get_model_feature_input(feature, object, context) | ||
| labels[feature] = label | ||
| results[feature] = kept_objects | ||
| v += [score, data] | ||
|
|
||
| return v, labels, results | ||
|
|
||
| def _get_model_feature_input(self, feature, object, context): | ||
| # context should include object | ||
| label, data = self.sm.label_feature(object, context, feature) | ||
|
|
||
| if feature == "color" or feature == "type": | ||
| score, kept_objects = self.elim_objects_discrete(context, label, feature) | ||
| else: | ||
| score, kept_objects = self.elim_objects_gradable(context, feature, label, data) | ||
|
|
||
| return label, score, data, kept_objects | ||
|
|
||
| def elim_objects_discrete(self, context, label, feature): | ||
| # we want to eliminate everything that the term label can NOT apply to | ||
| score = 0 | ||
| kept_objects = [] | ||
| for o in context.env: | ||
| this_label, data = self.sm.label_feature(o, context, feature) | ||
| if label == this_label: | ||
| kept_objects.append(o) | ||
| else: | ||
| score += 1 | ||
|
|
||
| return score, kept_objects | ||
|
|
||
| def elim_objects_gradable(self, context, feature, label, label_score): | ||
| # we want to eliminate everything that the term can NOT apply to | ||
| # that is, everything that the term fits LESS well than the target object | ||
| score = 0 | ||
| kept_objects = [] | ||
| for o in context.env: | ||
| this_label, data = self.sm.label_feature(o, context, feature) | ||
| if this_label == label and data >= label_score: | ||
| kept_objects.append(o) | ||
| else: | ||
| score += 1 | ||
|
|
||
| return score, kept_objects | ||
|
|
||
| def update_context(self, kept_objects): | ||
| return Context(kept_objects) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.