Chapter 1 / 6
✦ 0 XP
Chapter 1 / 6
Downsample the array
New here? Read the 60-sec briefing ↗The sensor feed is too dense to process at full rate. Pool each pair of readings down to the stronger one, halving the data while keeping the signal.
Pooling is a windowed reduce. Max-pooling halves resolution while keeping the strongest activation, the exact downsampling step inside a CNN.
YOUR TASK
- 1Each core owns one output cell covering inputs 2*i and 2*i + 1.
- 2Write the larger of the pair into out[i].
💡 Keep the strongest of each window: out[i] = max(inp[2*i], inp[2*i + 1]).
DEFINITIONS
- max-pooling
- — A downsampling step that slides a small window across the data and keeps only the largest value in each window, shrinking the array while preserving the strongest signal.
- downsampling
- — Reducing the resolution of a signal or feature map by collapsing groups of values into one, which cuts the work for every later layer.
- activation
- — A single value in a neural network's feature map, the output of one unit. Max-pooling forwards the strongest activation in each window.
core_11_pooling.mojo
PUZZLE 11 · POOLINGClear all 6 chapters to forge📡 Systems Core Shard