ICT Ministry Assistant Programmer Written Test: 2017প্রশ্ন ও সমাধান

Data communication is the exchange of data (in digital or analog form) between two devices via a transmission medium (wired or wireless), governed by an agreed set of rules (protocol). It requires five components: a message, a sender, a receiver, a transmission medium, and a protocol.

Based on the direction of data flow, communication is classified into three modes:

graph LR
 subgraph Simplex
 A1[Sender] -->|data only| B1[Receiver]
 end
 subgraph "Half-Duplex"
 A2[Device A] -->|"turn 1"| B2[Device B]
 B2 -->|"turn 2"| A2
 end
 subgraph "Full-Duplex"
 A3[Device A] -->|simultaneous| B3[Device B]
 B3 -->|simultaneous| A3
 end

Mode	Direction	Simultaneous?	Real-world example
Simplex	One-way only	N/A	Keyboard→CPU, TV/radio broadcast
Half-Duplex	Both ways, one at a time	No	Walkie-talkie, old-style hub Ethernet
Full-Duplex	Both ways at once	Yes	Telephone call, modern switched Ethernet

• Simplex wastes capacity if the reverse channel is never needed but is simplest and cheapest.
• Half-duplex uses the channel efficiently in both directions but needs a "handover" (e.g., "over" on a radio), which adds latency.
• Full-duplex gives the best throughput and lowest latency but generally needs separate transmit/receive paths (or techniques like echo cancellation) and costlier hardware.

Exam takeaway: The distinguishing test is "can both ends send at the same time?" — Yes = full-duplex, No but both directions possible = half-duplex, only one fixed direction ever = simplex.

Alpha-beta pruning is an optimization of the minimax algorithm: it returns the exact same move as plain minimax but skips branches of the game tree that cannot possibly change that decision.

Why minimax alone is slow
Minimax must visit every leaf to evaluate a game tree of branching factor b and depth d — an O(b^d) search. For real games (chess ~ b=35) this is infeasible beyond a few plies.

How pruning works
Two running bounds are passed down the recursion:

• alpha — the best value the Maximizer can guarantee on the path so far (starts at -infinity, only updated at MAX nodes).
• beta — the best value the Minimizer can guarantee on the path so far (starts at +infinity, only updated at MIN nodes).

At any node, once alpha >= beta, the remaining unexplored children are pruned — the opponent would never allow play to reach that branch, since a better alternative already exists elsewhere on the path.

graph TD
  A["MAX"] --> B["MIN"]
  A --> C["MIN (pruned after 1st child)"]
  B --> D(("3"))
  B --> E(("5"))
  C --> F(("2"))
  C -.->|"beta<=alpha, skip"| G(("not visited"))

Here B returns min(3,5)=3, so alpha at A becomes 3. While expanding C, its first child gives 2, making beta=2 at C; since beta(2) <= alpha(3), C can return at most 2 — worse than the 3 already found — so the remaining child of C is never explored.

Effect: with optimal move ordering, alpha-beta cuts the effective branching factor to the square root of b, reducing the search from O(b^d) to O(b^(d/2)) — letting the same time budget search roughly twice as deep.

Exam takeaway: alpha-beta pruning is exact (same decision as minimax), not approximate — it only removes provably irrelevant branches.