Generate Predictions and Analyze Accuracy
Since we have completed the training process, let’s actually try using it to predict the ‘wine quality’. For making predictions we’ll use the predict function of the model object. Let’s only give three examples as inputs and try to predict the wine quality for the 3.
Python3
# this will pass the first 3 rows of features# of our data as input to make predictionsmodel.predict(X_val.iloc[0:3, :]) |
Output:
array([[0.40581337],
[0.5295989 ],
[0.3883106 ]], dtype=float32)
Now, let’s compare our predictions with the target value.
Python3
y_val.iloc[0:3] |
Output:
0 0.4 9 0.4 12 0.4 Name: quality, dtype: float64
As we can see our predictions are pretty close to the real value i.e 0.4 in all three cases. You can define another function to convert the prediction into an integer to predict quality on a scale of 1 to 10 for better understanding, but that’s a trivial thing, the main thing is for you to understand the whole process depicted here.
Visualize Training Vs Validation Loss
You can analyze the loss and figure out if it is overfitting or not easy and then take appropriate measures accordingly.
Python3
loss_df = pd.DataFrame(losses.history)# history stores the loss/val# loss in each epoch# loss_df is a dataframe which# contains the losses so we can# plot it to visualize our model trainingloss_df.loc[:,['loss','val_loss']].plot() |
Output:
Key point to note while analyzing the accuracy/error for your model is :
Your loss function continuously decreasing, obviously, but that might not be the case for the validation dataset, and at some point, your model will overfit the data and the validation error will start increasing instead of decreasing. So, you can stop at the epoch where the validation loss seems to be increasing. You can also try some other optimization algorithms like early stopping (callback in Keras). You can read about it here.
Implementing Neural Networks Using TensorFlow
Deep learning has been on the rise in this decade and its applications are so wide-ranging and amazing that it’s almost hard to believe that it’s been only a few years in its advancements. And at the core of deep learning lies a basic “unit” that governs its architecture, yes, It’s neural networks.
A neural network architecture comprises a number of neurons or activation units as we call them, and this circuit of units serves their function of finding underlying relationships in data. And it’s mathematically proven that neural networks can find any kind of relation/function regardless of its complexity, provided it is deep/optimized enough, that is how much potential it has.
Now let’s learn to implement a neural network using TensorFlow
Contact Us