The worksheetDownload the PDF
Answers

8.7 Models as text

Learning · Robot club · about 15 min

BugBotLab

What this lesson is about

Saving and loading a model with JSON; models trained elsewhere.

Questions 6 marks in all

  1. [1 mark]What does this program print?

    import json
sample = ([16, 16], "wall-ahead")
back = json.loads(json.dumps(sample))
print(back)
print(back == sample)
    Answer:
    [[16, 16], 'wall-ahead']
False

    JSON has no tuples, so the pair comes back as a list, and a list is not equal to a tuple. That is why the models store samples as small dictionaries.

  2. [1 mark]What do json.dumps and json.loads do?

    1. Adumps turns data into text; loads turns text back into data
    2. Bdumps deletes a model; loads restores it
    3. Cdumps saves to a file; loads downloads from the internet
    4. DBoth turn text into data
    Answer: A. Text is what you can print, copy, paste into a file or send to a friend.

  3. [1 mark]What does this program print?

    import json
MODEL = '{"kind": "nearest-neighbour", "samples": [{"x": [16, 16], "label": "wall-ahead"}, {"x": [50, 50], "label": "open"}]}'
model = json.loads(MODEL)
data = [(s["x"], s["label"]) for s in model["samples"]]
print(model["kind"], len(data), data[1][1])
    Answer:
    nearest-neighbour 2 open

    The text becomes a dictionary. The two samples turn back into (features, label) pairs, and the second one's label is open.

  4. [1 mark]Why does the saved model include a kind field?

    1. ASo the program that loads it knows what to do with it
    2. BJSON refuses to save without one
    3. CIt records who made the model
    4. DIt makes the text shorter
    Answer: A. A nearest-neighbour model and a network need different code. The kind says which.

  5. [1 mark]How do you get a numpy weights table into a form JSON can write?

    1. AW1.tolist()
    2. Bjson.dumps(W1) works directly
    3. Cstr(W1)
    4. DW1.shape
    Answer: A. tolist() gives plain lists of numbers, and np.array(...) turns them back when you load.

  6. [1 mark]Which of these can JSON write down as they are?

    Tick every answer that is true.

    1. AA list of numbers
    2. BA dictionary
    3. CNone
    4. DA numpy array
    5. EA tuple, kept as a tuple
    Answer: A, B, C. JSON has lists, dictionaries, numbers, strings, True, False and None. Tuples become lists and numpy arrays need tolist() first.

The task: a model from text

Load the model in MODEL, classify the current view with it, and print model says: <label>. Do not drive.

# the two lines every program starts with: the commands, then the robot
from bugbot import *
connect()
import json
MODEL = '{"kind": "nearest-neighbour", "features": "depth grid rows 2 and 3", "samples": [{"x": [54, 52, 51, 51, 51, 51, 52, 54, 54, 52, 51, 51, 51, 51, 52, 54], "label": "open"}, {"x": [16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16], "label": "wall-ahead"}, {"x": [36, 47, 46, 46, 46, 16, 16, 16, 36, 47, 46, 46, 46, 16, 16, 16], "label": "gap-left"}, {"x": [16, 16, 16, 46, 46, 46, 47, 36, 16, 16, 16, 46, 46, 46, 47, 36], "label": "gap-right"}]}'
model = json.loads(MODEL)
print(model['kind'], len(model['samples']), 'samples')

The hint students can ask for: MODEL is a model saved as text. Load it with json.loads, classify the current view with it, and print model says: <label>. Do not drive.

A solution

from bugbot import *
connect()
def level_rows():
    return tof_grid()[16:32]               # rows 2 and 3: the 16 readings that look straight ahead

def nearest(sample, data):
    # the label of the recorded sample most like this one (smallest sum of squared differences)
    best_label, best_d = None, 1e18
    for feats, label in data:
        d = sum((a - b) ** 2 for a, b in zip(feats, sample))
        if d < best_d:
            best_label, best_d = label, d
    return best_label
import json
MODEL = '{"kind": "nearest-neighbour", "features": "depth grid rows 2 and 3", "samples": [{"x": [54, 52, 51, 51, 51, 51, 52, 54, 54, 52, 51, 51, 51, 51, 52, 54], "label": "open"}, {"x": [16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16, 16], "label": "wall-ahead"}, {"x": [36, 47, 46, 46, 46, 16, 16, 16, 36, 47, 46, 46, 46, 16, 16, 16], "label": "gap-left"}, {"x": [16, 16, 16, 46, 46, 46, 47, 36, 16, 16, 16, 46, 46, 46, 47, 36], "label": "gap-right"}]}'
model = json.loads(MODEL)
data = [(s["x"], s["label"]) for s in model["samples"]]
print("model says:", nearest(level_rows(), data))

Any program that meets the task's checks is marked correct in the simulator; this is one way, not the only way.