Praktikum 3

Install hnswlib terlebih dahulu.

!pip install hnswlib

Percobaan berikut akan membandingkan exact NN dengan HNSW pada 1000 data 2D.

import hnswlib
import numpy as np
import time
from sklearn.neighbors import NearestNeighbors

# ===========================
# 1. Buat data 2D acak
# ===========================
num_elements = 1000
dim = 2
data = np.random.random((num_elements, dim)).astype(np.float32)

# Query point
query = np.array([[0.5, 0.5]], dtype=np.float32)
k = 5  # cari 5 tetangga terdekat

# ===========================
# 2. Exact NN (Brute Force)
# ===========================
nn = NearestNeighbors(n_neighbors=k, algorithm='brute', metric='euclidean')
nn.fit(data)

start = time.time()
distances, indices = nn.kneighbors(query)
end = time.time()

print("=== Exact NN ===")
print("Indices:", indices)
print("Distances:", distances)
print("Waktu:", end - start, "detik")

# ===========================
# 3. HNSW
# ===========================
# Inisialisasi index HNSW
p = hnswlib.Index(space='l2', dim=dim)  

# Ukuran maksimum elemen yang bisa ditampung
p.init_index(max_elements=num_elements, ef_construction=100, M=16)

# Tambahkan data
p.add_items(data)

# Set parameter pencarian
p.set_ef(50)   # tradeoff speed vs accuracy

start = time.time()
labels, distances = p.knn_query(query, k=k)
end = time.time()

print("\n=== HNSW ===")
print("Indices:", labels)
print("Distances:", distances)
print("Waktu:", end - start, "detik")

Lakukan percobaan pada metric distance yang berbeda, 1000 vs 1jt data, 2D vs 5D data. catat hasilnya pada tabel yang anda buat sendiri seperti pada praktikum 1.