While loop with session state not giving wanted result

Using Streamlit
1

Hello everyone, here is my issue:
I’m doing an active learning app using streamlit and i’ve written the following code to do so:

import streamlit as st
from sklearn.preprocessing import MinMaxScaler
import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
from IPython import display
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler
from modAL.models import ActiveLearner
from modAL.uncertainty import uncertainty_sampling
from streamlit.state.session_state import SessionState


st.set_page_config(layout="wide")
@st.cache
def dataset_preprocessing(uploaded_files):
    
    # Import dataframe
    df = pd.read_csv(uploaded_files)
    
    # Clean useless data
    df.replace("", float("NaN"), inplace=True)
    df.dropna(inplace=True)

    # Get the name of all columns
    all_columns = list(df.columns.values)

    # Create the work dataframe
    transformed_df = df

    # Get columns that are not usable
    filtered_columns = transformed_df.dtypes[df.dtypes == np.object]
    list_of_columns = list(filtered_columns.index)

    # Transform those columns into usable one
    transformed_df[list_of_columns] = transformed_df[list_of_columns].apply(lambda col:pd.Categorical(col).codes)

    # Scale the two dataframes
    scaler = MinMaxScaler()

    transformed_df = scaler.fit_transform(transformed_df)

    #return the two dataframes
    return [df, pd.DataFrame(transformed_df,columns= all_columns)]



def training_loop(X_train,X_test,y_test,learner,quest):
    input_holder=st.empty()
    quest_holder=st.empty()
    x_holder=st.empty()
    plot_holder=st.empty()
    button_finish=st.button("Finish active learning")
    i=0
    if 'accuracy_score' not in st.session_state:
        st.session_state['accuracy_score'] =[0]
    if 'nb_queries' not in st.session_state:
        st.session_state['nb_queries'] =[0]
    while button_finish==False:

        fig, ax = plt.subplots()
        ax.plot(st.session_state['nb_queries'], st.session_state['accuracy_score'])
        plot_holder.pyplot(fig)

        X_train = pd.DataFrame(X_train)
        id_tolabel, x_to_label = learner.query(X_train) 
        quest_holder.write(quest)
        x_holder.write(df_transformed.iloc[id_tolabel])

        y = input_holder.text_input("Your answer",key=f"label_{i}")
        if y is not None:
            y_return=np.array([int(y)])
            learner.teach(x_to_label, y_return)
            X_train.drop_duplicates(inplace=True)
            X_train.drop(id_tolabel,axis=0,inplace=True)
            st.session_state['accuracy_score'].append(learner.score(X_test, y_test))
            st.session_state['nb_queries'].append(st.session_state['nb_queries'][-1]+1)
            st.write(X_train.shape)
            st.write(learner.score(X_test, y_test))
            st.write(st.session_state['nb_queries'])
            st.write(st.session_state['accuracy_score'])
            i+=1

    return learner

choice= st.sidebar.radio('Select one:', ['Normal dataset active learning','Plots for correlation'])
with st.expander("Your dataset, assuming that label is in last column" ,expanded=True):
        
    uploaded_files = st.file_uploader("Upload Dataset", type=["csv"], accept_multiple_files = False)
        
    if uploaded_files is not None:
        global df,df_transformed
        df,df_transformed=dataset_preprocessing(uploaded_files)
        df = pd.DataFrame(df)
        df_transformed=pd.DataFrame(df_transformed)
        global df_feature
        df_feature = df.drop(columns=df.columns[-1],axis=1)
        st.write(df_feature.head())
        st.write(df_transformed.head())

if choice == 'Normal dataset active learning':
    

    
    with st.expander("Labelling Function" ,expanded=False):
        
        quest = st.text_input("Label Question")
        
        if quest!="":
        
            X_train, X_test, y_train, y_test = train_test_split(df_feature, df_transformed[df_transformed.columns[-1]].values)

            scaler = MinMaxScaler()
            X_train = scaler.fit_transform(X_train)
            model = LogisticRegression()
            learner = ActiveLearner(
                estimator = model,
                query_strategy=uncertainty_sampling,
                X_training=X_train, y_training=y_train
            )
            training_loop(X_train,X_test,y_test,learner,quest)

if choice == 'Plots for correlation':
    for i in range (len(df.columns)-1):
        fig, ax = plt.subplots()
        ax.scatter(df.iloc[:,i], df.iloc[:,-1])
        ax.set_xlabel(str(df.columns[i]))
        ax.set_ylabel(str(df.columns[-1]))
        st.pyplot(fig)

The plotting part works fine, but the active learning part is not working as i wish. Indeed when i input new labels each time, it seems that for the first iteration everything works fine: accuracy is appended to the list, number of queries too. But when it comes to further iteration accuracies are being appended many times (2nd iteration=> accuracy appended 2 times, 3rd iteration=> accuracy appended 3 times).
I really couldn’t figure out what is going wrong in my code, also i’m not really used to session state so my use of it might be wrong.
Would be nice if someone can help me to fix this issue.
Thanks in advance.

2

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.