2ABT_behavior_models/plot_models_v_mouse.py at main · celiaberon/2ABT_behavior_models · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Wed Nov  3 13:20:48 2021

@author: celiaberon
"""

import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import os


def get_block_position_summaries(data):

    bpos = pd.DataFrame()

    for group_col in ['block_pos_rev', 'blockTrial']:

        block_start, block_end = [0,20] if group_col=='blockTrial' else [-20,-1]

        summary_stats = data.groupby(group_col).agg(
            phigh = pd.NamedAgg(column = 'highPort', aggfunc = 'mean'),
            phigh_std = pd.NamedAgg(column = 'highPort', aggfunc = 'std'),
            pswitch = pd.NamedAgg(column = 'Switch', aggfunc = 'mean'),
            pswitch_std = pd.NamedAgg(column = 'Switch', aggfunc = 'std'),
            n = pd.NamedAgg(column = 'Switch', aggfunc = 'count')).loc[block_start:block_end]

        summary_stats.index.name='block_pos'

        bpos=pd.concat((bpos,summary_stats))


    return bpos.reset_index()


def plot_by_block_position(bpos, subset='condition', **kwargs):

    sns.set(style='ticks', font_scale=1.6, rc={'axes.labelsize':18, 'axes.titlesize':18})

    color_dict=kwargs.get('color_dict', {key:val for key, val in zip(bpos[subset].unique(), np.arange(len(bpos[subset].unique())))})

    fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2, figsize=(10.5,3.5))
    ax1.vlines(x=0,ymin=0,ymax=1.05,linestyle='dotted',color='black')
    ax2.vlines(x=0,ymin=0,ymax=1 ,linestyle='dotted', color='black')

    for subset_iter in bpos[subset].unique():

        if type(color_dict[subset_iter])==np.int64:

            trace_color=sns.color_palette()[color_dict[subset_iter]]

            if subset_iter=='mouse':
                trace_color='gray'

        else:
            trace_color=color_dict[subset_iter]
        d = bpos.loc[bpos[subset] == subset_iter]

        ax1.plot(d.block_pos,d.phigh,label=subset_iter, alpha=0.8, linewidth=2, color=trace_color)
        ax1.fill_between(d.block_pos, y1=d.phigh - d.phigh_std / np.sqrt(d.n),
                                y2=d.phigh + d.phigh_std / np.sqrt(d.n), alpha=0.2, color=trace_color)

        ax1.set_yticks([0,0.5, 1.0])

        ax2.plot(d.block_pos,d.pswitch, label=subset_iter, alpha=0.8, linewidth=2, color=trace_color)
        ax2.fill_between(d.block_pos,y1=d.pswitch - d.pswitch_std / np.sqrt(d.n),
                                    y2=d.pswitch + d.pswitch_std / np.sqrt(d.n), alpha=0.2, color=trace_color)

        ax2.set_yticks(np.arange(0,0.6,step=0.1))#[0,0.1, 0.4])
    ax1.set(xlim=(-10,20), ylim=(0,1), xlabel='Block Position', ylabel='P(high port)')
    ax2.set(xlim=(-10,20), ylim=(0,np.max(bpos.pswitch)+0.05), xlabel='Block Position', ylabel='P(switch)')

    if len(bpos[subset].unique())<5:
        ax1.legend(loc=[0.5,-0.03], fontsize=16,frameon=False)
    sns.despine()
    plt.tight_layout()


def plot_scatter(df_mouse, df_model):

    sns.set(style='ticks', font_scale=1.6, rc={'axes.labelsize':18, 'axes.titlesize':18})
    sns.set_palette('dark')

    plt.figure(figsize=(4,4))
    plt.subplot(111, aspect='equal')
    plt.scatter(df_mouse.pswitch, df_model.pswitch, alpha=0.6, edgecolor=None, linewidth=0)
    plt.plot([0, 1], [0, 1], ':k')

    plt.xlabel('P(switch)$_{mouse}$')
    plt.ylabel('P(switch)')
    plt.xticks(np.arange(0, 1.1, 0.5))
    plt.yticks(np.arange(0, 1.1, 0.5))

    plt.tight_layout()
    sns.despine()


def plot_sequences(df, overlay=[], **kwargs):

    sns.set(style='ticks', font_scale=1.7, rc={'axes.labelsize':20, 'axes.titlesize':20})
    sns.set_palette('deep')

    overlay_label = kwargs.get('overlay_label', '')
    main_label = kwargs.get('main_label', '')
    yval = kwargs.get('yval','pswitch')

    df = df.astype('object') # to deal with histories being treated as categorical from sorting

    fig, ax = plt.subplots(figsize=(14,4.2))
    if len(overlay)>0:
        overlay = overlay.astype('object')
        sns.barplot(x='history',y=yval, data=overlay, label=overlay_label, color=sns.color_palette()[0], ax=ax, alpha=1.0)
        ax.errorbar(x='history',y=yval, yerr=yval+'_err', data=overlay, fmt=' ', label=None, color=sns.color_palette('dark')[0])

    sns.barplot(x='history',y=yval,data=df, color='k', alpha=kwargs.get('alpha',0.4), label=main_label, ax=ax, edgecolor='gray')
    ax.errorbar(x='history',y=yval, yerr=yval+'_err', data=df, fmt=' ', color='k', label=None)


    if len(overlay_label)>0:
        ax.legend(loc='upper left', frameon=False)
    ax.set(xlim=(-1,len(df)), ylim=(0,1), ylabel='P(switch)', title=kwargs.get('title', None))
    plt.xticks(rotation=90)
    sns.despine()
    plt.tight_layout()


def internal_prob(a, b, n):

    return np.nansum(a * b * n) / np.nansum(n)


def calc_confusion_matrix(df_mouse, col, df_model=None):

    #[[actual repeat * predict repeat, actual repeat * predict switch],
    #[actual switch * predict repeat, actual switch * predict swich]]
    # and can sub in right / left

    if df_model is None:
        df_model = df_mouse.copy()
    else:
        assert(np.all(df_mouse.history.values == df_model.history.values))

    N = df_mouse.n.values # same counts for model
    a = df_mouse[col].values
    b = df_model[col].values

    raw_confusion = np.array([[internal_prob(1-a, 1-b, N), internal_prob(1-a, b, N)],
                               [internal_prob(a, 1-b, N), internal_prob(a, b, N)]])

    norm_confusion = raw_confusion / raw_confusion.sum(axis=1)[:,np.newaxis]

    return norm_confusion


def plot_confusion(df, df_model, cm_fig=None, col='pswitch', color='Blues', seq_nback=3, delta=True):

    sns.set(style='white', font_scale=1.3, rc={'axes.labelsize':16, 'axes.titlesize':16})

    if cm_fig is None:
        cm_fig, ax = plt.subplots(nrows=1, ncols=1, figsize=(4.4, 2.2))
    else:
        ax = cm_fig.add_subplot(1, 2+delta, 2+delta)
        cm_fig.set_size_inches(8.4, 2.5)

    cm = calc_confusion_matrix(df, col, df_model)

    ax.imshow(cm, cmap=color)

    fmt='.2f'
    thresh = cm.max()/ 2.
    for i, row in enumerate(cm):

            for j, square in enumerate(row):

                ax.text(j, i, format(square, fmt),
                        ha="center", va="center",
                        color="white" if square > thresh else "black")

    column_dict = {'pswitch': ['repeat', 'switch'], 'pdecision':['left','right']}

    ax.set_xticks((0,1))
    ax.xaxis.tick_top()
    ax.xaxis.set_label_position('top')
    ax.set_xticklabels(('{} '.format(column_dict[col][0]),' {}'.format(column_dict[col][1])))
    ax.set_yticks((0,1))
    ax.set_yticklabels(('{}'.format(column_dict[col][0]),'{}'.format(column_dict[col][1])))
    ax.set(xlabel='predicted', ylabel='actual', ylim=(-0.5, 1.5))
    ax.invert_yaxis()
    plt.tick_params(top=False, pad=-2)
    plt.tight_layout()

    return cm_fig