The optimization of deep neural networks almost always relies on reducing the value of a loss function. For classification tasks, the loss function we used was cross entropy. Could we have instead just have used accuracy as the loss function? Why or why not? Hint: Think about what the loss function would look like

Answer using up to 3 sentences max.

Question

The optimization of deep neural networks almost always relies on reducing the value of a loss function. For classification tasks, the loss function we used was cross entropy. Could we have instead just have used accuracy as the loss function? Why or why not? Hint: Think about what the loss function would look like

Answer using up to 3 sentences max.

BlackTom AI · Accepted Answer

The question asks whether we could have used accuracy as the loss function for training deep neural networks on classification tasks, instead of cross-entropy, and to provide reasoning for the answer.
First, accuracy measures the proportion of correct predictions and is a discrete, non-differentiable metric that only changes when a prediction crosses the decision threshold (e.g., from incorrect to correct). This means it provides essentially no gradient information to guide optimization, so gradient-based methods cannot use it to adjust network weights.
Second, cross-entropy, by contrast, is smooth and differentiable with respect to the network’s outputs, yielding meaningful gradients at every training example that indicate how to adjust predictions to become more probable for the correct class. This enables effective optimization via backpropagation, whereas accuracy cannot drive learning in a fine-grained way.
Third, using accuracy as a loss would also create optimization dynamics that are non-smooth and often plateaus, making it difficult for the model to learn from intermediate states or handle mini-batch updates consistently. In short, accuracy is a useful evaluation metric but not a suitable objective for training neural networks; cross-entropy provides the necessary differentiable signal for gradient-based optimization.

CS-7643-O01, OAN, OSZ Quiz #2: Optimization of Deep Neural Networks (Lesson 3) and Convolution and Pooling Layers (Lesson 5)

View Explanation

Log in for full answers

Similar Questions

In a consumer society, many adults channel creativity into buying things

Economic stress and unpredictable times have resulted in a booming industry for self-help products

People born without creativity never can develop it

A product has a selling price of $20, a contribution margin ratio of 40% and fixed cost of $120,000. To make a profit of $30,000. The number of units that must be sold is: Type the number without $ and a comma. Eg: 20000

Which of the following statement regarding cost is correct:

Under the assumptions used in cost-volume-profit analysis, as the activity increases:

The CVP and break-even analysis are supported by a range of important assumptions. Which of the following is not an assumption of the CVP:

The break-even point is that at which:

More Practical Tools for Students Powered by AI Study Helper

Homework AI Solver

Stylized AI Paper Writer

Plagiarism Checker Assistant

Citation AI Academic Writing Tool

In-Class Translation Assistant

AI Note Generator

AI Quiz Answers

Past Exam Questions from University Test Bank

Smart Practice Assistant

Adaptive Practice

Making Your Study Simpler