{
"cells": [
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"import numpy as np\n",
"import pandas as pd\n",
"import matplotlib.pyplot as plt\n",
"import os\n",
"from hippocampus.analysis.daw_analysis import add_relevant_columns\n",
"import statsmodels.formula.api as smf\n",
"import seaborn as sns\n",
"from tqdm import tqdm_notebook as tqdm\n",
"\n",
"from definitions import RESULTS_FOLDER\n",
"from hippocampus.environments import HexWaterMaze"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"en = HexWaterMaze(6)"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"res_dir = os.path.join(RESULTS_FOLDER, 'mb_spatialmemory')\n",
"\n",
"params = pd.read_csv(os.path.join(res_dir,'params.csv'))\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"params.hist()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"df = pd.read_csv(os.path.join(res_dir, 'spatial_agent0'))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"df.columns"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"df.head()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"df.plot(x='total trial', y='escape time')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"data = df[['trial', 'total trial', 'session', 'escape time','platform', 'state']]"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"n_agents = 19"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"def get_first_trial_info(data):\n",
" d2 = data.pivot_table(index='total trial')\n",
" d2['previous platform'] = d2['platform'].shift(1)\n",
" first_trials = d2[d2['trial']==0]\n",
" first_trials = first_trials.drop(0).pivot_table(index='session')\n",
" return first_trials"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"def get_surrounding_states(state, env, rec_depth=2):\n",
" surrounding_states = [state]\n",
" for i in range(rec_depth):\n",
" added_states = []\n",
" for s in surrounding_states:\n",
" neighbours = np.flatnonzero(env.adjacency_graph[s])\n",
" for n in neighbours:\n",
" if n not in surrounding_states and n not in added_states:\n",
" added_states.append(n)\n",
" surrounding_states += added_states\n",
" return surrounding_states\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"df = pd.read_csv(os.path.join(res_dir, 'spatial_agent{}'.format(1)))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"def get_allo_index(agent_data, env):\n",
" \"\"\"Get the allocentricness index, defined as the amount of time spent around the previous platform location during\n",
" first trials of sessions.\n",
"\n",
" :param agent_data:\n",
" :param env:\n",
" :return:\n",
" \"\"\"\n",
" first_trials = get_first_trial_info(agent_data)\n",
" prop_times = []\n",
" for ses in range(1, 11):\n",
" states = np.sort(agent_data[(agent_data.session == ses) & (agent_data.trial == 0)]['state'])\n",
" previous_platform = first_trials['previous platform'][ses]\n",
" surrounding_states = np.sort(np.array(get_surrounding_states(int(previous_platform), env)))\n",
"\n",
" time_spent = np.isin(states, surrounding_states).sum()\n",
" prop_times.append(time_spent)\n",
"\n",
" return np.mean(prop_times)\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"scores = []\n",
"\n",
"data = []\n",
"\n",
"for a in tqdm(range(n_agents)):\n",
" df = pd.read_csv(os.path.join(res_dir, 'spatial_agent{}'.format(a)))\n",
" data.append(df)\n",
" #ft = get_first_trial_info(df)\n",
" scores.append(get_allo_index(df, en))\n",
" "
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"get_first_trial_info(data[0])"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"# load model-based data"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"def get_model_weights(data):\n",
" add_relevant_columns(data)\n",
" data['Stay'] = data['Stay'].astype('int')\n",
" data = data[['Stay', 'PreviousReward', 'PreviousTransition']]\n",
" mod = smf.logit(formula='Stay ~ PreviousTransition * PreviousReward', data=data)\n",
" res = mod.fit()\n",
" model_based_weight = -res.params['PreviousTransition[T.rare]:PreviousReward']\n",
" model_free_weight = res.params['PreviousReward']\n",
" return model_based_weight, model_free_weight\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"weights_mb = []\n",
"for a in range(n_agents):\n",
" df = pd.read_csv(os.path.join(res_dir, 'twostep_agent{}'.format(a)))\n",
" mb_weight, mf_weight = get_model_weights(df)\n",
" weights_mb.append(mb_weight)\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"allocentric_scores = []\n",
"for a in tqdm(range(n_agents)):\n",
" df = pd.read_csv(os.path.join(res_dir, 'spatial_agent{}'.format(a)))\n",
" allocentric_scores.append(get_allo_index(df, en))\n",
"\n",
"mb_scores = []\n",
"for a in range(n_agents):\n",
" df = pd.read_csv(os.path.join(res_dir, 'twostep_agent{}'.format(a)))\n",
" mb_weight, mf_weight = get_model_weights(df)\n",
" mb_scores.append(mb_weight)\n",
"\n",
"allocentric_scores_lesion = []\n",
"for a in tqdm(range(n_agents)):\n",
" df = pd.read_csv(os.path.join(res_dir, 'spatial_agent{}_lesion'.format(a)))\n",
" allocentric_scores_lesion.append(get_allo_index(df, en))\n",
"\n",
"mb_scores_lesion = []\n",
"for a in range(n_agents):\n",
" df = pd.read_csv(os.path.join(res_dir, 'twostep_agent{}_lesion'.format(a)))\n",
" mb_weight, mf_weight = get_model_weights(df)\n",
" mb_scores_lesion.append(mb_weight)\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"plt.scatter(np.log(allocentric_scores), mb_scores, c=sns.cubehelix_palette(len(scores)))"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"plt.figure(figsize=(4,4))\n",
"sns.regplot(allocentric_scores, mb_scores)\n",
"\n",
"plt.ylabel('Model based index')\n",
"plt.xlabel('Allocentricness score')"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"colpal = sns.color_palette()"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"fig, ax = plt.subplots()\n",
"\n",
"plt.xlim([4.5,0])\n",
"plt.ylim([-1,4.5])\n",
"sns.regplot(np.log(allocentric_scores), mb_scores,color=colpal[1])\n",
"sns.regplot(np.log(allocentric_scores_lesion), mb_scores_lesion, color=colpal[4])\n",
"\n",
"plt.ylabel('Model based index')\n",
"plt.xlabel('Allocentricness score')\n",
"ax.spines['right'].set_visible(False)\n",
"ax.spines['top'].set_visible(False)\n",
"\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"score_data = pd.DataFrame({})\n",
"score_data['model based'] = np.concatenate([mb_scores, mb_scores_lesion])\n",
"score_data['allocentric'] = np.concatenate([np.log(allocentric_scores), np.log(allocentric_scores_lesion)])\n",
"score_data['group'] = ['control'] * n_agents + ['lesion'] * n_agents\n",
"\n",
"\n",
"\n",
"sns.lmplot(y='model based', x='allocentric', data=score_data, hue='group', palette=[colpal[1], colpal[4]])\n",
"\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": [
"sns.palplot(sns.color_palette())"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.6.8"
}
},
"nbformat": 4,
"nbformat_minor": 2
}