Convolution and pooling geometry

Tutor me on https://mlprep.iwase.dev/item/ml-screen-convolution-standard-shape/. This is a screening item. If window.mlPrepAgent is available, read attempts for item ml-screen-convolution-standard-shape before tutoring. Inspect my attempt, keep the item ID, and do not reveal the full answer first. After a real attempt, append its record with recorded_by agent, include agent_session_id when available, and read it back.

Problem

The input shape is $28\times28\times3$ . A convolution uses $10$ filters of size $3\times3$ , stride $2$ , padding $1$ , and one bias per filter. The convolution is followed by $2\times2$ max pooling with stride $2$ and no padding. Compute the convolution shape, convolution parameter count, and pooled shape.

Reveal answer or reference solution

Each convolution dimension is

\left\lfloor\frac{28+2(1)-3}{2}\right\rfloor+1=14,

so the convolution shape is $14\times14\times10$ . The parameter count is

3\cdot3\cdot3\cdot10+10=280.

Pooling gives $7\times7\times10$ and adds no parameters.

Previous attempts

Loading attempts saved in this browser…

Review Screening Results

Convolution and pooling geometry

Problem

Previous attempts

Appears in